The Burden of Voting in the 2019 Canadian Federal Election
Chapter 5: How much does the burden matter?
We have seen that the various perceived burdens of voting tend to be relatively small but that there are exceptions, especially among youth, Indigenous electors living on reserves, electors with severe disabilities, and regular abstainers. But there is still the possibility that a small burden has a big impact on the decision to vote, especially if the benefit of voting is relatively small. As Aldrich (1993) elegantly argued, a small cost may make a big difference in a low-cost low-benefit context.
Our goal in this chapter is to estimate the extent to which the intention to vote is affected by the various burdens that we have examined in the previous chapters. In other words, how would turnout be affected if all electors expected registering, going to the polling station, casting a ballot, and deciding whom to vote for to be very easy?
To do so we must take into account the other, mainly motivational, factors that shape turnout: interest in politics, belief that voting is a duty, and partisan affiliation. Following Blais and Daoust (2020), we focus on two powerful motivational attitudes: interest in politics and the belief that voting is a civic duty. There is strong empirical evidence that each of these two attitudes affects the decision to vote, that is, those who are not very interested in politics or who do not feel that they have a moral obligation to vote are much more inclined to abstain. We include a third variable, whether one feels close to any political party, as those who feel close to a party are more prone to vote (Campbell et al. 1960).
The level of political interest is measured by a straightforward question asking how interested in politics one is (Q49, Wave 1 and 3); mean interest on a 0 to 1 scale is .71; the most common answer is "somewhat interested", followed by "very interested." footnote 25 Sense of civic duty is tapped by the question proposed by Blais and Achen (2019) whereby respondents were invited to tell if they deem voting to be a duty or a choice (Q55r, Wave 1); 74% chose duty. Party identifiers are defined as those who indicate that they feel close to any federal political party (Q64, Wave 1); 45% of the respondents feel close to a party.
The question is thus how much of the decision to vote or abstain is shaped by these motivational factors and how much, by the obstacles. Our approach is the following. Our dependent variable is the intention to vote, measured in wave 1 of the questionnaire (Q35, Wave 1). That variable takes the value of 1, if the person responded that they were certain to vote, which was the case for 80% of the sample and 0 otherwise (respondents who said they were likely, unlikely, or certain not to vote). We also replicate the analyses with reported turnout in the post-election survey (Q37, Wave 3), in which 90% of the respondents claim to have voted. The results are reported in Appendix D, Table 1.
The official turnout in the election was 67%. footnote 26 The survey clearly overestimates turnout, as is common in surveys about electoral participation. There are two main reasons for that overestimation. The first is self-selection. People who are more interested in politics (and who are more prone to vote) are more likely to respond to "political" surveys. Attrition accentuates this in the case of panel surveys, as those who are less interested in politics are more likely to quit in the second and third waves. The self-selection bias is thus greater in wave 3, and this is why most of our analyses are performed with wave 1 data. The second reason is social desirability: some people are reluctant to admit that they do not vote (because the "good" citizen is supposed to vote). The reported vote question in the post-election wave was framed in such a way as to make it easier for people to indicate that they did not vote this time (Morin-Chassé et al. 2017), but this does not completely eliminate the social desirability bias.
Even though existing research shows that there is a bias in the distribution of the dependent variable, the available evidence suggests that this may not substantially affect the relationships that we observe. Achen and Blais (2016) use the American National Election Studies (ANES) to examine the correlates of intention to vote, reported vote, and validated vote. They find that "all of the substantively relevant variables are statistically significant and with the correct sign in all three equations" (page 200); and that "our findings support a circumscribed and qualified endorsement of the current practice of conflating studies of intended vote, reported vote, and validated vote." (page 207)
Using OLS estimations, we predict the likelihood of being certain that one will vote, using the three motivational variables and the four burdens, while also controlling for socio-demographic characteristics. The full results are reported in Appendix D, Table 2. Figure 5.1 presents the findings for motivations and burdens for model 1 in Table 2. We can see that burdens do matter: each of the four burdens has an independent effect on the propensity to vote. However, the motivational factors (interest in politics and duty) have a much greater impact. Whether one is interested in politics and whether one feels that they have a moral obligation to vote are more consequential than whether voting is perceived to be easy or difficult.
Among the four burdens, the one with the biggest coefficient is registration. Everything else being equal, the probability of being certain to vote decreases by 13 points when registering is perceived to be very difficult. That being said, this burden does not apply to the great majority of Canadians who are already registered, and, for them, this is not a concern, as there is nothing they need to do in order to register. Our data suggest that turnout would increase by one percentage point if registration was perceived to be very easy by everyone. This estimate is obtained by multiplying .08 (the mean of the registration burden) by .125 (the coefficient of the registration burden). Using the same approach, we estimate that turnout would increase by 4 points if the decision burden was completely absent, 3 points if there was no perceived difficulty associated with going to the polling station, and 2 points if everybody thought that casting a ballot was very easy.
Getting rid of the decision burden does not appear to be a realistic option, especially considering that many Canadians are not very well informed about the parties, leaders, and candidates. Our data suggest that, if we focus on the procedural aspects, those aspects that Elections Canada can possibly improve, the maximum increase in turnout, if every aspect of the procedure were to be perfectly smooth, is six percentage points. footnote 27 We would venture to say that the realistic potential total gain is probably a few percentage points. This is not a lot, but our data suggest that it is possible to increase turnout marginally, especially by making it easier for everyone to go to the polling station and cast their ballot.
It makes sense to believe that these burdens do not have the same impact on everybody. There are some people who are willing to vote whatever the difficulty, because they are passionate about politics and/or believe that it is their civic duty to vote. Others may be ambivalent and be tempted to abstain if they face any kind of obstacle. This suggests that the various burdens will have a greater effect on the turnout decision among those who are less interested in politics or who think that voting is a personal decision, not a duty. This interpretation is supported in Figure 5.2, which shows the impact of the "polling station" and "casting a ballot" burden depending on one's level of interest and sense of civic duty (see Table 2 in Appendix D for the full results). We can see that the impact of these burdens is much greater among those who are not at all interested in politics and among those who do not have a sense of civic duty; in this case, the presence of an obstacle provides people with an extra reason to stay home.
We have noted in Chapter 4 that the various burdens tend to be higher among youth. This raises the question of whether the lower turnout that is observed among younger citizens is due to burdens. To determine whether this is the case, we first regress intention to vote on our age categories, with and without controls. The full results are shown in Appendix E. There is to start with a 15-percentage point gap between the turnout of the 18–24 group and the 45–54 group (our reference category). When we control for other socio-demographic characteristics, especially for the fact that those aged 18–24 tend to be poorer, unmarried and more mobile, that gap is reduced to 10 points. What is most interesting is the fact that the gap is not weaker when we introduce motivational factors. Among youth, burdens appear to matter more than the lack of motivation.
Our last task is to determine whether the greater burdens experienced by some groups lead to a lower turnout. We proceed in the following way: We first ascertain whether turnout is indeed lower in that group. We then introduce socio-demographic characteristics, the motivational variables and the burdens to see if they account for the lower turnout. Finally, we determine whether there is evidence that the various burdens have a bigger impact on turnout in that group.
The full findings are presented in Appendix F, Tables 1 to 4, for each of our groups and subgroups: Indigenous people on or off reserve, those with moderate or severe disability, those who were born after 1998 or acquired citizenship after the previous 2015 election and for whom this is the first federal election for which they had the right to vote, regular abstainers who had never voted before, and finally the NEET group.
Figure 5.3 shows the gap in intended turnout between members of these groups and the rest of the sample without any controls. As expected, turnout is systematically lower in all of the groups. The gap is, however, moderate in most groups, except two, where it is really big (about 25 points for NEET and Indigenous people on reserve), and one where it is even bigger (52 points for regular abstainers). In the case of NEET, that gap disappears when we control for socio-demographic characteristics. footnote 28
Let's focus on the two groups for which there is a big turnout gap, which cannot be explained as a pure reflection of socio-demographic characteristics: Indigenous people living on reserves and regular abstainers. Figure 5.4 shows how the turnout gap is reduced for these two groups, as we introduce socio-demographic characteristics, then motivational variables, and finally the burdens. Let us start with Indigenous people living on reserves. It is important to keep in mind that the turnout gap is quite low for Indigenous people living off reserve compared with the gap among those living on reserve. Figure 5.4 indicates that controlling for the burdens decreases the turnout gap by only 3 points. We conclude that it is not mainly because of the voting burdens that Indigenous people living on reserves are less prone to vote.
What about regular abstainers? Figure 5.4 shows that the most important factor consists of motivations. They are less interested in politics and less inclined to believe that they have a civic duty to vote, and this explains in good part their abstention. We can see, however, that the turnout gap for this group decreases by 7 points when the burdens are taken into account. Clearly, for this group as for most citizens, motivation (or the lack of) matters more but burdens come into play as well. At least some of them would be willing to vote if they were convinced that voting is quite easy.
The concern is that members of these groups are more strongly affected by the various burdens when they have to decide whether to vote or not. We test that possibility in column 5 of Tables 1 to 4 in Appendix F. If the burdens have a greater impact for a given group, the interaction variable should be significant and negative, meaning that it has a bigger negative effect in that group. There are a total of 32 interaction variables included in Tables 1 to 4 (column 5), and only five of them have a significant negative coefficient. The various burdens are higher in those groups and turnout is lower, but the burdens are only a small part of the explanation for the lower turnout, and they do not matter more than in the rest of the electorate.
Let us examine one final piece of evidence about the role of burdens in the decision not to vote. Wave 3 (post-election) of the survey asked those who indicated that they did not vote to tell the main reason why they abstained (Q38, Wave 3). The questionnaire proposed a long list of 18 possible reasons, five of them referring to lack of motivation (04 to 08) such as lack of interest and eight process-related ones (09 to 16) such as long lineups. All in all, 44% chose a lack of motivation reason and only 11% a burden. This is consistent with our verdict that lack of motivation matters more than burdens though the latter also play a role.
Our analyses suggest that burdens matter, not a lot, but still significantly. According to our estimates, turnout would increase by 10 percentage points if everybody found it very easy to register, make up their mind about who to support, go to the polling station, and cast a ballot. This is, of course, unrealistic, especially with respect to the decision costs. We have thus focused on the other three "procedural" costs, about which Elections Canada has some leverage. Turnout could be increased by six points if these burdens were completely abolished. We believe that a more realistic goal would be to substantially reduce these burdens, and that this could boost turnout by a few points.
We also find that these burdens have a greater impact on the propensity to vote among those who are less interested in politics and who have a weaker sense of civic duty. These people may be more difficult to reach, and it may be a greater challenge to inform them that voting may be easier than they think.
As for specific subgroups, we have identified two where turnout is particularly low: Indigenous people living on reserves and regular abstainers. The lower participation of Indigenous people living on reserves is a complex issue that we cannot fully address in this study. Our data suggest, however, that burdens are not the main reason for this lower participation. As for regular abstainers, lack of motivation is the principal factor, but perceived burdens also play a role.
Footnotes
Return to source 25 The response categories are: not at all (0), not very (.33), somewhat (.66) and very (1).
Return to source 26 https://www.elections.ca/content.aspx?section=res&dir=rec/eval/pes2019/vtsa&document=index&lang=e
Return to source 27 Three points associated with the polling station burden, two points with the ballot burden, and one point with registration.
Return to source 28 This reflects the fact that turnout is generally lower among youth and among the unemployed. There is nothing special about being both young and unemployed.