Does Voting in One Election Reduce the Expected Cost of Voting in Subsequent Elections?

Abstract The decision to vote is partly based on the expected cost of voting. We test the hypothesis that voting in one election reduces the expected cost of voting in the following election, as voters learn that the cost of voting is low. Using three different datasets—the National Electors Study conducted during the 2019 Canadian federal election; a two-wave YouGov survey in British Columbia and Quebec in 2008 and 2009, at the time of the federal and subsequent provincial elections; and a five-wave survey conducted for the Making Electoral Democracy Work project in Bavaria in 2013 and 2014, before and after the Land, federal and European elections—we find that voters who voted in a previous election perceive it will be easier to vote in a subsequent election. We also find evidence that voting leads to more accurate estimates of how little time it takes to vote.

takes to decide which candidate to vote for and how much time it takes to go to a polling station, wait in a queue, and cast a vote. In this study we ascertain how easy or difficult citizens perceive voting to be. We use three datasets: the National Electors Study in Canada, an original dataset from British Columbia and Quebec, and the Making Electoral Democracy Work project. We rigorously test our hypothesis and find evidence that citizens who vote in one election are more accurately able to estimate the costs it will take for them to vote in a subsequent election, even after controlling for a host of other variables, including motivational factors and duty.
The decision to vote or not hinges at least partly on the evaluation of the potential benefits of voting versus the costs, which are associated with the time and energy needed to go to the polling station and cast a vote, as well as to obtain information about the candidates and the voting process (Downs, 1957;Riker and Ordeshook, 1968;Blais, 2000). Participating in elections is a fundamental mechanism through which citizens hold their elected officials accountable (Achen and Bartels, 2017;Schaltegger and Torgler, 2007). Yet many countries are experiencing a decline in political participation (Kostelka and Blais, 2021), a topic that has been well explored in scholarship focused on who votes and why (Campbell, 2006;Leighley and Nagler, 2014;Powell, 1982;Verba et al., 1995;Wolfinger and Rosenstone, 1980). There is also a vast literature on the impact that making voting easier and more accessible has on turnout (Holbein and Hillygus, 2016;Bhatti, 2012;Brady and McNulty, 2011;Dyck and Gimpel, 2005;Gimpel and Schuknecht, 2003;Garmann, 2017;Potrafke and Rösel, 2018).
There is considerable evidence that those who vote in an election are more likely to vote in subsequent elections. Many scholars have interpreted this finding as evidence for habit formation (Green and Shachar, 2000;Denny and Doyle, 2009;Meredith, 2009;Dinas, 2012;Fujiwara et al., 2016;Plutzer, 2002); this is the idea that voting becomes automatic. The validity of the habit formation hypothesis has also been questioned (Bhatti et al., 2016;Bechtel et al., 2018;Blais and Daoust, 2020;Hernaes, 2019;Jessen et al., 2021).
In this research note, we propose and test an alternative interpretation: voting in one election makes people perceive it to be easier to vote in the following election. The hypothesis is based on the assumption that for most people voting is easy and also that there is a learning mechanism, as voting makes people more familiar with the voting process. This is what practising is about. The more you undertake an activity or a task, the easier that activity or task gets over time. The more you bike, the easier biking becomes; the more often you see an object, the easier it becomes to track it (Pinto et al., 2010). In short, people learn from doing something, so that doing it the second time is (a bit) easier than the first. As far as we can tell, this simple hypothesis has never been tested in previous research.
In this study, we rigorously test if voting in one election makes voters perceive it will be easier to vote in the subsequent one. We use three different data sets, and we show compelling evidence that, even after controlling for a host of other factors, people who voted in the previous election are more likely to say in the next election that voting is easy. This is an important finding, as our study suggests that the perceived cost of voting in future elections is reduced by actually casting a vote.

Research Design and Results
We use three different datasets to test the hypothesis that voting in one election makes it easier to perceive voting to be easy in a subsequent election. In all these analyses, the dependent variable is the perceived easiness/difficulty of voting (in Study 2 and Study 3, our dependent variable is the difference between the perceived cost in two successive elections) 1 and the main independent variable is whether the person voted in the previous election. 2 In our final analysis, we compare people's expectations of the time they expect voting to take with the actual time it took, and we show that those who had voted in the previous election had relatively accurate expectations while those who had not voted were prone to overstate the time it would take. All the models we present are based on ordinary least squares (OLS) estimations.
In Study 1, we use the National Electors Study (NES), conducted online in Canada with eligible electors in the context of the October 2019 Canadian federal general election (see Blais and Sevi, 2021). This is a panel survey that consists of three waves: before the election period, during the election campaign and after the election. We use the first wave, which includes many questions about the cost of voting and has a large sample size of approximately 42,000 respondents. The respondents are representative of the Canadian population, based on province, age and gender.
In wave 1, Canadians were asked how easy or difficult they thought it would be to decide which candidate to vote for, to go to the polling station, and to vote once arrived at the polling station. The response categories are very easy, somewhat easy, somewhat difficult, and very difficult (see online Appendix 1 for exact question wordings). These variables were recoded from 0 (very easy) to 1 (very difficult). The mean for each of these variables is .31, .12 and .11, respectively, which indicates that the various costs tend to be low.
Respondents are then asked to estimate how much time, in minutes, they thought it would take to go to the polling station and then cast a vote once arrived. The mean for going to the polling station and casting a vote is 14 and 12 minutes, respectively. All in all, most people think it will not take much time to exercise their right to vote.
We have five dependent variables, which correspond to five indicators of the cost of voting, related to the anticipated time and difficulty associated with going to the polling station and then casting a vote, as well as the challenge of deciding which candidate to vote for. The main independent variable is whether the respondent voted in the previous federal election held in October 2015. We include many controls: age; age squared; gender; language spoken at home; foreign born; education; and motivations of voting, such as interest in politics, civic duty and party identification. There is evidence that highly motivated citizens are prone to believe that voting is easy (Blais and Daoust, 2020). Table 1 presents the results. For each of the five indicators, those who had voted in the previous federal election expected the cost of voting to be smaller. The differences are statistically significant and relatively large. Let us take the cost associated with going to the polling station. Everything else being equal, the predicted difficulty of going to the polling station is .21 among those who did not vote in the previous election, and only .10 among those who did vote; having voted in the previous election reduces the perceived future cost by more than half. Similarly, the expected time it would take to go to the polling station is predicted to be 13 minutes among those who had voted previously and 21 among those who did not, a relative reduction of 38 per cent.
These results strikingly support our hypothesis that voting in one election reduces the perceived costs of voting in the following election. Our data are only cross-sectional, however. 3 Therefore, we supplement these findings with panel data which allow us to directly observe whether the fact of voting in one election does lead people to expect it to be easier in the following election-that is, data that trace perceived costs over two successive elections and which show that voting in the first election leads to reduced expected costs in the second election. This is precisely the data that we have in Study 2 and Study 3.
In Study 2, we draw on an original dataset from a two-wave internet survey conducted by YouGov in the provinces of British Columbia and Quebec. The first wave was conducted during the last week of the October 2008 Canadian federal election, and the second wave took place at the time of the subsequent provincial election, which occurred in December 2008 in Quebec and in May 2009 in British Columbia. 4 Similar to Study 1, the dependent variables are how easy or difficult respondents thought it would be to go to the polling station and make up their mind about the parties and party leaders. Distinct from Study 1, we calculate the dependent variable by subtracting the cost at the provincial level (the second election) from that at the federal level (the first election). The response categories are very easy, somewhat easy, somewhat difficult, and very difficult. Responses to the two "making up their mind" questions are strongly correlated (.84) and were thus combined. We have two indicators of cost: the relative difficulty of deciding and going to the polling station. The variables are recoded on a 0 (very easy) to 1 (very difficult) scale.
We have two measures of these perceived costs for the same respondents, the first (in wave 1) just before the federal election and the second (in wave 2) just before the following provincial election. The dependent variable is the change in perceived cost between the federal and the provincial election, and thus ranges from −1 to +1. We also know (from wave 2) whether the respondent did or did not vote in the first (federal) election. Our strategy is to see if having voted in the federal election made people revise downward the expected costs of voting in the subsequent provincial election. In doing so, we are tapping into changes in perceptions that took place between the two elections. We also incorporate controls for sociodemographic and motivational variables, as in Study 1.
The results are presented in Table 2. Having voted in the federal election is indeed significantly associated with a reduction in the difficulty of deciding and of going to the polling station. The impact here is smaller, which is what we should expect since the panel data allow for a more conservative and reliable estimate of the effect of having voted. That being said, the effect is still meaningful. Having voted in the previous election reduces the perceived cost of deciding whom to vote for and of going to the polling station by 6 per cent and 4 per cent respectively. 5 Study 1 and Study 2 are based on Canadian data. We would, of course, like to know whether the pattern observed in Canada holds elsewhere. We have been able to find one other study conducted outside Canada with panel data like Study 2 that includes a question that taps respondents' perceptions of the cost of voting in subsequent elections. This allows us to determine whether having voted in one election makes people revise downward the difficulty of voting in the following election. This is Study 3.
In Study 3, we use data from the Making Electoral Democracy Work (MEDW) project. MEDW includes many studies, but one of them (and only one) is a panel survey conducted in Bavaria that covers three successive elections: the Land and the federal elections that were held (one week apart) in September 2013 and the European elections that took place in May 2014. The survey has five waves: the first was in the week preceding the Land election, the second in the week following the Land election and preceding the federal election, the third in the week following the federal election, the fourth in the week preceding the European election, and the fifth (which we do not use) after the European election. We perform three analyses. We determine whether having voted in the Land election reduced the perceived costs of voting in the following federal and European elections and whether having voted in the federal election had a similar impact in the European elections.
The dependent variables are the difference in how easy or hard respondents thought it would be to vote in the federal and Land elections, European and Land elections, and European and federal elections. The response categories are very easy, somewhat easy, somewhat hard, and very hard, and they are recoded on a 0 to 1 scale. The main independent variables are whether the respondent voted in the previous Land or federal election, and we control for the ex ante perceived cost of voting as well as for socio-demographic and motivational variables.
The results are presented in Table 3. The "having voted in the previous election" variable is negative in all three instances and significant in two of them. If we take the two elections that took place just one week apart, our results suggest that having voted in the first (Land) election led to a 10 per cent decline in the expected cost of voting in the following federal election. 6 Finally, we go back to the NES (2019 Canadian election) data to compare how much time citizens expect it will take to vote and how much time it actually takes to vote. Note that we can only consider those who voted in the 2019 election (those who did not vote did not go to the polling station and were thus not asked how much time it took). We can distinguish, among those 2019 voters, those who did and did not vote in the previous 2015 election. Our prediction is that those who had not voted in the 2015 election tended to exaggerate the time it would take. Our prediction is confirmed. Those who had not voted in the 2015 election indicated, before the election, that it would take 23 minutes, on average, to go to the polling station and 16 minutes to cast their vote once arrived. In the post-election survey, the same respondents revealed that it actually took, on average, 10 and 8 minutes, respectively. All in all, these two tasks took 21 minutes less than expected. Note that those who had voted in 2015 also overestimated the time it would take by 4 minutes for going to the polling station and 5 minutes for casting their ballot. But their overestimation was much smaller: 9 minutes overall, compared to 21 minutes among those who had not voted in 2015. Having voted in the previous election contributed to a more accurate estimate of the low amount of time that voting requires. People learned that it would not take much time to cast their vote.

Conclusion
In this research note, our goal has been to test the hypothesis that citizens who vote in one election find it easier to vote in a subsequent election. We have used three different data sets, two of them based on panel surveys, and we have found consistent evidence that those who vote in an election revise downward their expectations about the cost of voting in the following election. We also show that voting leads to a more accurate estimate of how little time it takes to vote. The findings are consistent with a learning mechanism. 7 Voting in one election helps people realize that voting is (mostly and relatively) easy. The fact that voting in one election is associated with voting in the next election is often interpreted as meaning that voting becomes a habit. This interpretation has been questioned in recent research (Blais and Daoust, 2020;Jessen et al., 2021;Dunaski, 2021). In the case of Brazil, for instance, Dunaski (2021: 7) finds "no evidence that compulsory voting is habit-forming." Our results suggest another interpretation. Someone who votes in one election is likely to find that voting is easier than she thought, and this makes her a bit more prone to vote in a subsequent election.
Our study pertains to only two countries, Canada and Germany, and more research is needed to determine if this pattern holds more generally. But the available evidence suggests that in most countries, voting is quite easy (see Blais and Daoust, 2020). The challenge, therefore, is to convince new electors and previous abstainers that voting is easier than they think. 8 Supplementary Material. To view supplementary material for this article, please visit https://doi.org/10. 1017/S0008423922000063.

Notes
1 We are interested in subjective expectations about how easy or difficult it will be to vote, as measured before the election, since it is those subjective perceptions that drive the decision to vote or not (Blais et al., 2019). 2 Achen and Blais (2016) show that the relationship between reported voting and different independent variables is similar to that with validated voting. We recognize, however, that effects for reported turnout in this study may be slightly overestimated. It is also possible that there is a self-selection bias in who takes surveys, a limitation that we acknowledge (Dahlgaard et al., 2019). 3 The NES is a panel dataset with three waves, but we do not have data on the perceived costs of voting in two different elections. As such, we use only the first wave. 4 Since we are using panel data, our sample is much lower than in the original first wave sample. As can be seen in online Appendix 2, however, the socio-demographic profile of the panel respondents does not differ much from that of the original sample. 5 More precisely, keeping all the other variables constant, the predicted cost of deciding is .49 among those who voted in the previous election, compared to .52 among those who did not vote. The equivalent numbers are .50 and .52 for the cost of going to the polling station. 6 Keeping all the other variables constant, the predicted cost of voting is .46 among those who voted in the previous election, compared to .51 among those who did not vote. 7 We acknowledge that there may be other reasons why those who vote in one election are more prone to vote in a subsequent election. Our study indicates only that learning that voting is easy is one reason. 8 Note that we have not dealt here with the cost of registration because in the two countries covered in this study (Canada and Germany), registration is state initiated. More research is required to determine if the cost of registration is overestimated in countries where registration is an individual's responsibility.