open Secondary menu

Canadian attitudes towards voting during the COVID-19 pandemic – Wave 2

Report prepared for Elections Canada

Peter John Loewen & Eric Merkley
PEARL (Policy, Elections, and Representation Lab)
Munk School of Global Affairs and Public Policy
University of Toronto

August 17, 2020

Introduction

This report considers the impact of the COVID-19 pandemic on voter turnout and voting site administration in Canada. We build on our previous report to examine how the pandemic is shaping Canadians' intention to turnout, and to understand more deeply how and why different actions taken by Elections Canada might support voter turnout in an election conducted during a pandemic.

Voting in Canada overwhelmingly occurs in person, and there are important reasons to maintain this practice as much as possible, not least generating confidence in the electoral process, maintaining past practices to which both parties and voters have adapted their behaviour, and encouraging the civic action of voting alongside other citizens. This report pays particular attention to understanding how on-site voting measures can generate confidence in the voting process and can support the intention to vote.

Our report has five high-level findings:

  • Finding 1: The COVID-19 pandemic continues to depress voter turnout intention. This is unchanged from our previous report. The effect of COVID-19 on depressing turnout is especially pronounced among those who perceive a high level of risk from COVID-19.
  • Finding 2: We probe a number of different forms of voting – in-person on election day, advance voting, postal voting, etc. In the aggregate, we find that preferences are stable and largely unchanged from mid-June. As found in our previous report, voters are willing to vote via in-person methods, but a large minority would prefer voting by mail. This method is not opposed by a large share of voters.
  • Finding 3: Respondents support mandatory mask usage at voting stations. While this is majority-supported, a text analysis of reasons for opposition to this suggest that those who suppose this do so for reasons of personal liberty, rather than inconvenience. Risk perceptions of COVID-19 is a major predictor of support for the mandatory usage of masks.
  • Finding 4: A majority of Canadians would support weekend voting, and another quarter are unsure of their opinion on this policy. There is not significant opposition to weekend voting. Risk perceptions of COVID-19 is a major predictor of support for weekend voting.
  • Finding 5: We probed the effect of five different safety measures on perceptions of safety of in-person voting and on likelihood of voting in person. All five measures – limiting the number of people in a polling place, enforcing social distancing, mandating mask usage, providing hand sanitizer, and providing single-use pencils –have positive effects on perceived safety and the likelihood of voting in person. We also find substantial cumulative effects, suggesting that each additional measure has a positive effect.

In what follows, we review the methodology of our survey, followed by a detailed presentation of our findings. We conclude with policy implications.

Methodology―Survey Design

Our study relies on a survey conducted among 2,499 adult Canadians between August 10 and August 12, 2020. Our sample was provided by Dynata Inc, an international survey sample provider who control a large number of proprietary survey panels. Our sample was constructed with national level quotas for age and gender and provincial population quotas. We then weight the data using an iterative proportional fitting package (ipfweight) in STATA16.1 Respondents completed the survey on the Qualtrics platform.

Our survey is principally focused on COVID-19. The survey also contains a battery of questions on past political participation, leader evaluations, and the like. The survey is non-partisan, and no elements concerning partisan preferences have been analyzed for this report or shared with Elections Canada. In addition to COVID-19 related questions, respondents were also asked to answer a number of questions designed in consultation with Elections Canada.
In the first section of the survey, respondents were asked the following, where some respondents were primed to think about COVID-19 and others were not.

  1. Considering the current situation with COVID-19, if an election was held tomorrow, how likely or unlikely would you be to vote?

    Or

    If an election was held tomorrow, how likely or unlikely would you be to vote?

    • Certain to vote
    • Likely to vote
    • Unlikely to vote
    • Certain not to vote

In the final section of the survey, respondents were asked 5 questions. Respondents were then randomized to one of the following two questions:

  1. Now imagine a federal election was to be held in the coming weeks. If you were to vote, and considering the current situation with COVID-19, what voting method do you think you would use? [select only one]

    Or

    Now imagine a federal election was to be held in the coming weeks. If you were to vote, and considering the current situation with COVID-19, which of the following voting methods would you definitely NOT be willing to use? [select all that apply]

    • At the polling station on election day
    • At the advance polling station
    • At a local Elections Canada office before election day
    • By mail
    • Other (please specify)
    • Don't know yet

They were then asked their agreement with the following statements:

  1. "All electors should be required to wear a mask to enter a polling station and vote"
  2. "Polling day should be held on Saturday and Sunday (2 days), rather than on a Monday"
    • Strongly agree
    • Somewhat agree
    • Somewhat disagree
    • Strongly disagree
    • Don't know

And if they indicated they disagreed with either statement they were asked to explain why in an open text box

    • Why do you disagree with requiring all electors to wear a mask to enter a polling station and vote?
    • Why do you disagree with moving polling day to Saturday and Sunday?

Following the standard suite of questions respondents were given a conjoint experiment. First, they received a brief introduction:

In the following section we are going to provide you with several hypothetical scenarios related to the administration of in-person voting in the next federal election. Please read carefully. We will ask you some questions about each scenario.

Then they were given three scenarios where the voting safety measures were randomized:

Imagine the following measures were in place for in-person voting in the next federal election:

  • [Strict limits on the number of people allowed into a polling station at a time/No limits on the number of people allowed into a polling station].
  • [Social distancing measures to keep people at least 2 meters apart/No social distancing measures in place].
  • [Requirement to wear a mask when inside the polling station/No requirement to wear a mask while inside the polling station].
  • [Hand sanitizer is provided/Hand sanitizer is not provided].
  • [Single-use pencils are provided/Single-use pencils are not provided].

After each scenario, they were asked the following:

  1. How safe or unsafe do you think it would be to vote in person at a polling station with these measures in place?
    • Very unsafe
    • Somewhat unsafe
    • Somewhat safe
    • Very safe
  2. How likely would you be to vote in person at a polling station with these measures in place?
    • Very likely
    • Somewhat likely
    • Not very likely
    • Not at all likely

The remainder of our survey is available upon request. Variable descriptions can be found in Appendix A.

Results

Frequencies

We begin our analysis by considering the "topline" responses to each of our questions. Table 1 presents our respondents' self-estimated likelihood of voting in the next federal election. The first two columns present the results of these questions asked at the beginning and the end of the survey in the previous wave. The second two columns show the results of primed and non-primed respondents in this wave. The COVID-19 prime marginally reduced people's reported likelihood of voting (p<0.1, one-tailed), but this effect is less precisely estimated than in wave 1 owing to the loss of statistical power from the between-subjects design.

Table 1. Self-reported likelihood of voter turnout with and without COVID-19 prime
Wave 1 Wave 2
Vote likelihood at start of survey without prime Vote likelihood at end of survey with prime Vote likelihood without prime Vote likelihood with prime
Certain to vote 73.4 66.9 75.1 71.8
Likely to vote 19.1 22.3 17.2 19.5
Unlikely to vote 4.6 7.6 4.7 6.4
Certain not to vote 2.9 3.3 3.1 2.5
N 2,497 2,497 1,207 1,293

When we asked half of respondents about a preferred voting method, we find substantial preferences for non in-person voting methods. These results are shown in Table 2, with the first two columns presenting the results from the previous wave. While the majority of individuals indicate that they would vote in person, either at a polling station (29.4%), at an advance polling station (28.6%), or at an Elections Canada office before election day (5.1%), a substantial share of individuals (21.8%) indicate that they would prefer to vote by mail. Just one-in-eight respondents (12.9%) indicate that they do not know how they would vote.

When we instead asked half of respondents about methods they definitely would not use, we find that 33% indicate they would not vote in a polling station on election day, 24% would not vote in advance at a polling station, 17% indicate they would not vote at an Elections Canada office, while 20% indicate that they would not vote by mail. For the most part, then, voters are not substantially opposed to any particular method of casting a ballot, but a concerning number indicate an unwillingness to vote on election day or using another in person method. These results have remained stable between the two waves fielded so far.

Table 2. Respondent preferences over voting method
Wave 1 Wave 2
Intend to use Definitely not use Intend to use Definitely not use
Election day 29.2 33.1 29.4 32.5
Advance polls 28.5 22.9 28.6 23.6
Elections Canada office 4.7 17.0 5.1 16.8
Mail 23.2 19.1 21.8 20.0
Other 2.2 1.7 2.2 2.2
Don't know 12.2 30.5 12.9 35.1
N 1,243 1,257 1,265 1,234

Table 3. Agreement with mandatory masks and weekend voting
  Masks Weekend voting
Strongly agree 70.4 37.1
Somewhat agree 17.0 25.0
Somewhat disagree 4.8 9.5
Strongly disagree 3.6 5.0
Don't know 4.3 23.5
N 2,499 2,499

We finally ask respondents their level of agreement with statements signaling support for the mandatory usage of masks in polling stations and voting over the weekend rather than on a Monday. The results are shown in Table 3. A remarkable 70% of respondent strongly agree with the mandatory usage of masks, with only 8% reporting any disagreement. In contrast, support for weekend voting is softer and less certain as 24% of respondents reported that they didn't know whether they agreed with the position or not while 37% strongly agree and 15% report some level of disagreement.

In-depth analysis

COVID-19 Priming Experiment

Canadians are overwhelmingly concerned about the COVID-19 pandemic, and, as we demonstrated in our previous report, this concern is lowering their confidence in the safety of in-person voting. This has the potential to reduce voter turnout in the future if the COVID-19 pandemic remains top-of-mind for Canadians.

Table 4. Heterogeneous effects of COVID-19 prime on turnout intention, OLS estimates
  Coef. SE Coef. SE
Prime 0.02 0.04 0.07* 0.04
COVID-19 risk 0.02** 0.01 0.02 0.01
COVID-19 risk * Prime -0.01 0.01 0.00 0.01
Education 0.13*** 0.03 0.13*** 0.03
Education * Prime -0.05 0.04 -0.04 0.04
Age 0.16*** 0.02 0.15*** 0.02
Age * Prime -0.02 0.03 -0.01 0.03
Location 0.01 0.02 0.00 0.02
Location * Prime 0.04 0.03 0.05* 0.03
Female -0.01 0.01 -0.02 0.01
Female * Prime -0.03* 0.01 -0.02 0.01
COVID-19 risk perception 0.09*** 0.02
Risk perception * Prime -0.10*** 0.03
Constant 0.51*** 0.03 0.47*** 0.03
R2 0.07 0.07
N 2493 2493

Note: * p<0.1; ** p<0.05; *** p<0.01. All covariates scaled from 0-1.

In our previous report, we show that respondents reduce their reported likelihood of voting after considering the COVID-19 pandemic for 15 minutes (i.e. the length of the survey dedicated in large part to COVID-19). We found that this effect was strongest among older respondents and people concerned about COVID-19. Here, we implement a different between-subjects design to test the effects of COVID-19 on turnout intention. One group of respondents got a variant of the turnout intention that asked them to consider the COVID-19 pandemic (the treatment prime), while another group did not. We find a small 1-point difference on a 0-1 scale between the two groups in the expected direction, which is marginally significant (p<0.1, one-tailed).

We test for heterogeneous effects on turnout intention by estimating an OLS regression model where we interact the treatment with relevant covariates: gender, urban location, age, education, and COVID-19 risk. We estimate a second model where we also interact the treatment with our risk perception index. The estimates are shown in Table 4. Model 1 finds little evidence of heterogeneous effects by age, as we found in the previous report, but women are more responsive to the treatment prime (p<0.1). There is no predicted treatment effect among men, but a 2-point reduction in turnout intention for women (p<0.1).

We do find, however, strong evidence that the treatment prime reduces vote intention among those who perceive a high degree of risk from COVID-19 (p<0.01). Our models predict that these respondents reduce their reported intention to vote by almost 4 points in response to the treatment prime, consistent with the results from the previous report.

Improving the Perceived Safety of In-Person Voting

The previous survey wave indicated that a majority of Canadians perceive in-person voting to be somewhat or very unsafe. We also found that people who perceive voting to be unsafe – and who were very concerned about COVID-19 – were more likely to reduce their intention to vote in response to thinking about the COVID-19 pandemic. Above, we find similarly that priming people with the COVID-19 pandemic reduces intention to vote, primarily among those who perceive a high level of risk associated with COVID-19.

Consequently, it is essential to discover ways to improve the perceived safety of in-person voting. In the previous wave we asked respondents whether they believed a series of measures would improve the safety of in-person voting: 1) controls on the numbers of people allowed into a polling station; 2) social distancing measures to keep people at least 2 meters apart; 3) requiring everyone to wear a mask; 4) single-use pencils; 5) provision of hand sanitizer; 6) early voting hours. Respondents were generally of the opinion that all of these measures would improve safety, ranging from 61% of respondents for single use pencils to 83% of respondents for social distancing measures. However, we do not get a sense from this question how much each of these measures affects one's perceptions of in-person voting safety or their own willingness to vote in such a situation.

For this survey, we designed a conjoint experiment to overcome these problems. Conjoints are a tool often used in market research to allow for a more fine-grained understanding of people's preferences when the choice situation is complex and multidimensional, such as in-person voting when multiple different packages of safety measures are possible.

We begin by instructing respondents that they will be provided with several possible scenarios related to the administration of in-person voting:

In the following section we are going to provide you with several hypothetical scenarios related to the administration of in-person voting in the next federal election. Please read carefully. We will ask you some questions about each scenario.

We then provide them with three such scenarios where each measure is randomized:

Imagine the following measures were in place for in-person voting in the next federal election:

  • [Strict limits on the number of people allowed in a polling station at a time/No limits on the number of people allowed into a polling station].
  • [Social distancing measures to keep people at least 2 meters apart/No social distancing measures in place].
  • [Requirement to wear a mask when inside the polling station/No requirement to wear a mask while inside the polling station].
  • [Hand sanitizer is provided/Hand sanitizer is not provided].
  • [Single-use pencils are provided/Single-use pencils are not provided].

After each scenario respondents were asked: "How safe or unsafe do you think it would be to vote in person at a polling station with these measures in place?" (Very safe, somewhat safe, not very safe, not at all safe); and "How likely would you be to vote in person at a polling station with these measures in place?" (Very likely, somewhat likely, not very likely, not at all likely).

We estimate the effects of each measure on perceptions of safety and vote likelihood in a pair of models. The estimates can be interpreted as the causal effect of each attribute averaging across the distributions of all other attributes in the model. This will tell us which of these measures improves safety perceptions and the likelihood of in-person voting the most.

The estimates are provided in Figure 1. All measures significantly improve perceptions of safety and the likelihood of in-person voting at the p<0.001 level, but there are some differences in effect sizes. The effects of single use pencils (6 points each on 0-1 scales) and hand sanitizer (9 and 8 points, respectively) are smaller than the effects of limits on the number of people, social distancing, and mandatory masks, which range from 12 to 14 points. These are large effects.

Figure 1. Effects of safety measures on perceived safety and the likelihood of in-person voting. Note: 95% confidence intervals.

Figure 1: Effects of safety measures on perceived safety and the likelihood of in-person voting

Text version of "Figure 1: Effects of safety measures on perceived safety and the likelihood of in-person voting"

This difference of means plot (above) is used to identify: the impact of proposed safety measures on electorsí perceived safety of voting in-person during the Covid-19 Pandemic (plotted as a circle), and the likelihood of electors voting in-person (plotted as a square).

The Y-axis presents the variables (single-use pencils, sanitizer provided, mandatory mask, social distancing numerical limits).

The x-axis presents a 0-1 scale.

Note: 95% confidence intervals.

  Number of points for proposed
safety measures on electors' perceived safety of voting in-person (0-1 scale)
Number of points for the likelihood of electors voting in-person (0-1 scale)
Proposed safety measures    
Single-use pencils 6 6
Sanitizer provided 9 8
Mandatory mask 12 14
Social Distancing 11 12
Numerical Limits 13 14

The above analysis assumes that each measure has independent effects on safety perceptions and the likelihood of in-person voting. However, there may be interactive effects as well. It is possible that respondents see these measures as fully or partially substitutable, such that there are diminishing marginal returns to additional safety measures. Or, respondents may see these measures as complimentary to one another, such that the effect of a measure is even higher when additional measures are in place.

We do not have theoretical expectations for which measures are more or less likely to exhibit interactive effects. Consequently, to test this possibility we create a continuous measure ranging from 0-5 for the number of safety measures respondents were exposed to in a given scenario. We examine the effect of this index on safety perceptions and include a quadratic term in order to account for non-linearity. The predictions from this model are shown in Figure 2.

Figure 2. Effect of safety measures on perceived safety (left) and the likelihood of in-person voting (right). Note: 95% confidence intervals.

Effect of safety measures on perceived safety (left) and the likelihood of in-person voting (right)

Text version of "Effect of safety measures on perceived safety (left) and the likelihood of in-person voting (right)"

These two scatter plots (above) present a quadratic relationship between:

  • The number of safety measures used (x-axis, a scale from 1 to 5) and the predicted perceived safety (y-axis, a scale from 0 to 1), and
  • The number of safety measures used (x-axis, scale from 1 to 5) and the likelihood of electors voting in person (y-axis, scale from 0 to 1).
  0 safety measures 1 safety measure 2 safety measures 3 safety measures 4 safety measures 5 safety measures
Predicted perceived safety (scale from 0 to 1) 2 points 3 points 4 points 5 points 6 points 8 points

  0 safety measures 1 safety measure 2 safety measures 3 safety measures 4 safety measures 5 safety measures
Predicted likelihood of electors voting in-person (scale from 0-1) 3 3.5 4 5.5 6.5 8

As is apparent in the figure, the effects of safety measures are increasing in the number of safety measures – rather than decreasing – for both safety perceptions and the likelihood of in-person voting.  The effect of moving from no measures to 1 measure is only 5 points, while it is 17 points when moving from 4 measures to 5. Moving from scenarios with 0 safety measures to all 5 increases safety perceptions from 0.24 on the 0-1 scale to 0.79. Similarly, the effect of moving from no measures to 1 safety measure is 5 points in the likelihood of in-person voting, while it is 15 points when moving from 4 to 5 safety measures. Moving from scenarios with 0 safety measures to all 5 increases the likelihood of in-person voting from 0.30 on the 0-1 scale to 0.81. In short, there are gains to having many safety measures in place.

We can also explore whether there are important group differences in the influence of safety measures on safety perceptions and in-person voting likelihood. We test this by interacting relevant covariates: COVID-19 risk, age, education, urban location, and gender, with our safety measure index. We estimate a second set of models that includes a similar interaction with COVID-19 risk perceptions. The estimates are provided in Appendix B. Table 5 displays the predicted means at relevant values of each covariate for those exposed to scenarios with no safety measures and those exposed to scenarios with all five.

Table 5. Predicted perceptions of voting safety and likelihood of in-person voting by values on treatment and relevant covariates
Safety Perceptions In-Person Voting
# of safety measures None All Difference None All Difference
Model 1
Female Yes 0.08 0.72 0.64*** 0.16 0.74 0.58***
No 0.24 0.72 0.48*** 0.31 0.75 0.44***
Age 35 0.25 0.70 0.45*** 0.30 0.73 0.43***
65 0.04 0.75 0.71*** 0.14 0.77 0.63***
Urban Small town 0.12 0.73 0.61* 0.19 0.75 0.56
Large city 0.19 0.72 0.53* 0.25 0.74 0.49
Education Secondary 0.23 0.71 0.48*** 0.29 0.72 0.43***
Undergrad 0.11 0.73 0.62*** 0.19 0.76 0.57***
COVID-19 risk Yes 0.15 0.71 0.56 0.23 0.74 0.51
No 0.17 0.73 0.56 0.24 0.75 0.51
Model 2
Risk perception High 0.06 0.67 0.61* 0.11 0.71 0.60***
Low 0.38 0.85 0.47* 0.48 0.83 0.35***

Note: * p<0.1; ** p<0.05; *** p<0.01 denote differences in treatment effect between groups.

We describe the effects below as moving from scenarios with no safety measures to those with all five. Women are more likely to increase their safety perceptions than men (64 vs. 48 points; model 1, column 3), which is strongly significant (p<0.01). People aged 65 are expected to increase their safety perceptions by 71 points, compared to only 45 points for those aged 35. This difference is also significant (p<0.01). Individuals with university education are expected to increase their safety perceptions 62 points, compared to only 48 points for those with high school education (p<0.01). Individuals residing in small towns increase their perceptions of safety more so than those in large cities, though this difference is smaller (p<0.1). There is no relationship between perceived COVID-19 health risk and the perception of the safety of measures proposed.

We also see that people who perceive high levels of risk from COVID-19 increase their safety perceptions by 61 points compared to only 47 points for those who perceive little risk (p<0.1; model 2, column 3). We see virtually identical patterns when looking at respondents' reported likelihood of in-person voting.

In short, the adoption of safety measures is more effective in improving safety perceptions and voting likelihood among women, the elderly, the highly educated, and among those who perceive high levels of risk from COVID-19. The end result is that the mean differences in safety perceptions and the likelihood of in-person voting become smaller in the presence of these safety measures.

Table 6. OLS estimates, support for mandatory masks and weekend voting
Mandatory masks Weekend voting
1 2 1 2
COVID-19 risk 0.01
(0.01)
-0.02*
(0.01)
0.02
(0.01)
-0.01
(0.01)
Education 0.19***
(0.03)
0.17***
(0.03)
0.10***
(0.03)
0.08**
(0.03)
Income 0.03
(0.03)
0.03
(0.03)
-0.01
(0.03)
-0.01
(0.03)
Age 0.25***
(0.03)
0.22***
(0.03)
-0.02
(0.03)
-0.05
(0.03)
Location 0.03
(0.02)
-0.01
(0.02)
0.00
(0.02)
-0.02
(0.02)
Female 0.05***
(0.01)
0.04***
(0.01)
-0.01
(0.01)
-0.01
(0.01)
Quebec 0.01
(0.02)
0.06***
(0.02)
0.00
(0.03)
0.03
(0.03)
Ontario 0.00
(0.02)
0.02
(0.02)
-0.01
(0.03)
-0.00
(0.03)
West -0.02
(0.02)
-0.00
(0.02)
-0.01
(0.03)
-0.00
(0.03)
COVID-19 risk perception   0.35***
(0.03)
  0.23***
(0.03)
Voted in 2019 0.03**
(0.02)
0.06***
(0.02)
Constant 0.59*** 0.36*** 0.65*** 0.48***
R2 0.07 0.17 0.01 0.04
N 2269 2269 2269 2269

Note: Standard errors in parentheses; * p<0.1, ** p<0.05, *** p<0.01. All covariates re-scaled from 0-1 for their minimum and their maximum

Determinants of Support for Mandatory Masks and Weekend Voting

Almost 88% of respondents somewhat or strongly agree with masks being mandatory at polling stations, while 62% agree with weekend voting at some level. Our data allows us to evaluate which groups of Canadians are more supportive or hostile to these measures. We estimate two models for each question. The first predicts support for the measure with demographic characteristics such as region, gender, age, location of residence, 2018 household income, education, and whether they (or someone in their household) are personally at risk of complications from COVID-19. The second model includes those same characteristics, while adding COVID-19 risk perceptions (0-1 index) and whether or not the respondent voted in the 2019 federal election (1=Yes). Each of our dependent variables are scaled from 0-1 where 1 indicates respondents who are strongly supportive of the measure. "Don't knows" are coded as a neutral point.

Model 1 shows that education (p<0.01), age (p<0.01), and gender (p<0.01) are strongly associated with support for mandatory masks. Someone with a university level education would be expected to have support levels 8 points higher than someone who only graduated high school. An individual aged 65 would be expected to have support levels 10 points higher than someone aged 35, and women score 5 points higher in their support than men. Interestingly, there is no association between COVID-19 health risk and support for mandatory masks when accounting for other factors in the model. Model 2 shows that COVID-19 risk perceptions are strongly associated with support for masks (p<0.01), along with past voting history (p<0.05). Voters in 2019 are 3 points more supportive of mandatory masks compared to non-voters, while people with the highest perception of COVID-19 risk are 35 points more supportive than those who have the lowest perception of risk.

Support for weekend voting is not as well predicted by our model. Model 3 shows that education is the only demographic determinant of support for this measure (p<0.01). Someone with a university level education would be expected to have support levels 4 points higher than someone who only graduated high school. Model 4 highlights significant associations between COVID-19 risk perceptions (p<0.01) and voting history (p<0.01), on the one hand, and support for this measure, on the other. 2019 voters are 6 points more supportive of weekend voting, while people who perceive the most risk from COVID-19 are 23 points more supportive of this measure than those who perceive the least risk.

And what exactly is behind opposition to mandatory masks and weekend voting? We mined the open-text questions with the automated content analysis program Quanteda. After processing the data, we generate word frequencies to shed light on common themes. Among responses to the mask open-text question where the word was composed of 5 characters or more, the most consistent theme that emerges in this analysis are related to personal choice, freedom, and rights. There appears to be a separate dimension related to questions about the scientific evidence in favour of masks.

The data shows the same results for the open-text weekend voting question. The 4-character words show that work, lack of time, and busyness is a dominant consideration for our respondents. The 5-character words show more clearly the considerations of importance. One consideration appears to be the interference of weekend voting with family time and religious worship, while another is concerned with the crowds and traffic on the weekend.

Determinants of Voting Method Preferences

COVID-19 concern and perceptions of election safety are also closely tied to voting methods people are willing to use or avoid. We estimate two models predicting respondents' preference for each of the following voting methods: on election day, at an advance poll, at an Elections Canada office, and by mail, as well as their opposition to each of those methods. The first model includes demographic characteristics such as region, gender, age, location of residence, 2018 household income, education, and whether they (or someone in their household) are personally at risk of complications from COVID-19. The second model includes those same characteristics, while adding level of COVID-19 concern and whether or not the respondent voted in the 2019 federal election.

The estimates of our models are provided in Appendix C, while predicted probabilities are shown in Table 5 to illustrate the substantive importance of the effects since logistic regression coefficients are not directly interpretable. We find that women are 8 points less likely to choose election day than men (Table 7, model 1, column 1; p<0.01), while COVID-19 risk perception is negatively related with this choice (model 2, column 1). People who have the highest risk perceptions are 14 points less likely to vote on election day than people who have the lowest risk perception (0.26 vs. 0.40; p<0.05). People who voted in 2019 are 18 points more likely to choose election day (p<0.01).

When asking respondents what voting methods they would definitely not use, younger people and those at risk of COVID-19 complications appear more inclined to oppose election day voting (model 1, column 5). Those at risk of COVID-19 complications are 7 points more likely to oppose this voting method (0.37 vs. 0.30; p<0.05), while individuals aged 35 are expected to be 5 points more likely to oppose this voting method than someone aged 65 (0.35 vs. 0.30; p<0.1).  Respondents who perceive the most risk of COVID-19 are 10 points more likely to report they would definitely not vote on election day than those who perceive the least amount of risk (0.36 vs. 0.26; p<0.1; model 2, column 5). Individuals who voted in 2019 were also 10 points more likely to oppose voting on election day (0.36 vs 0.26; p<0.05; model 2, column 5), likely an indication of stronger preferences for voting method by likely voters.

Table 7. Predicted probability of supporting or opposing voting method by values on relevant covariates
Support Oppose
Election day
1
Advance poll
2
EC office
3
Mail
4
Election day
5
Advance poll
6
EC office
7
Mail
8
Model 1
Female Yes 0.26*** 0.27 0.04 0.25*** 0.32 0.20*** 0.17 0.22
No 0.34*** 0.30 0.06 0.19*** 0.34 0.28*** 0.17 0.20
Age 35 0.31 0.24*** 0.07*** 0.21 0.35* 0.27*** 0.19** 0.20
65 0.29 0.34*** 0.03*** 0.24 0.30* 0.20*** 0.14** 0.22
Urban Small town 0.28 0.27 0.05 0.21 0.36 0.22 0.15 0.20
Large city 0.31 0.29 0.05 0.23 0.31 0.26 0.18 0.22
Income Low 0.30 0.27 0.04* 0.25* 0.34 0.24 0.17 0.21
High 0.30 0.30 0.07* 0.20* 0.32 0.24 0.17 0.21
Education Secondary 0.31 0.29 0.07** 0.17*** 0.33 0.26 0.18 0.20
Undergrad 0.30 0.28 0.04** 0.25*** 0.32 0.23 0.17 0.22
COVID risk Yes 0.28 0.28 0.07** 0.24 0.37** 0.27* 0.19 0.20
No 0.31 0.29 0.04** 0.21 0.30** 0.22* 0.16 0.22
Model 2
COVID Risk perception High 0.26** 0.28 0.05 0.31*** 0.36* 0.23 0.17 0.24*
Low 0.40** 0.29 0.07 0.08*** 0.26* 0.27 0.17 0.14*
Voted in 2019 Yes 0.32*** 0.31*** 0.05 0.21*** 0.34** 0.25* 0.16 0.20
No 0.14*** 0.11*** 0.05 0.34*** 0.24** 0.19* 0.21 0.23

Note: * p<0.1; ** p<0.05; *** p<0.01.

Older respondents are modestly more inclined to select advance polls (p<0.01; model 1, column 2), and less inclined to oppose them (p<0.1; model 1, column 6). Those aged 65, for example, are 10 points more likely to vote at advance polls than those aged 35 (0.34 vs. 0.24), while they are 7 points less likely to avoid them (0.20 vs. 0.27). They are also 4 points less likely to vote at an Elections Canada office than those aged 35 (p<0.01; model 1, column 3), and 5 points more likely to definitely not use that voting method (p<0.05; model 1, column 7). People at risk of COVID-19 complications are 3 points more likely to choose the Elections Canada office (p<0.05; model 1, column 3), and 5 points more likely to avoid advance polls (p<0.1; model 1, column 6). COVID-19 risk perceptions are not significantly associated with choosing or avoiding either advance polls or voting at an Elections Canada office (model 2).

Finally, there are important demographic differences in willingness to vote by mail as shown in Table 6. Women are 6 points more likely than men to say they would use mail-in ballots (0.25 vs. 0.19; p<0.01; model 1, column 4). Higher income respondents are 5 points less likely to vote by mail than those with lower income (0.20 vs. 0.25, p<0.1; model 1, column 4), while university educated respondents are 8 points more likely to vote by mail than those with only secondary school education (0.25 vs. 0.17; p<0.01; model 1, column 4). COVID-19 risk perceptions are associated with voting by mail in the expected direction, with people with the highest level of risk perceptions being 23 points more likely to vote by mail than those with the lowest level (0.31 vs. 0.08; p<0.01; model 2, column 4). Oddly, risk perceptions are associated with opposition to vote by mail, though this effect is only marginally significant (p<0.1).

In short, respondents' concern about COVID-19 and perceptions of election safety are powerful determinants as to whether or not respondents will choose to vote in person – especially on election day – or whether to vote by mail.

Discussion

The COVID-19 pandemic continues to upend daily life. Importantly, it continues to act as a drag on individuals' expectations that they will participate in any upcoming election. With disruption comes substantial innovation by organizations of all types. Understanding how those who interact with organizations will react to various adaptations is key to maintaining necessary activity during the pandemic. Election management is not excepted from this requirement. As our report shows, there are a number of innovations which electoral management bodies can undertake to increase both confidence in the safety of voting and in turnout intention. Importantly, the effects of these measures are cumulative, suggesting that more extensive efforts at improving the safety of voting will yield greater effects.

We wish to underline our continued finding that the most important predictor of willingness to engage in voting in light of COVID remains individual-level perceived risk of COVID. Those who perceive themselves and others to be at risk, independent of their other demographic characteristics, are the most likely to respond positively to various safety measures and to adjust their turnout intention accordingly. Communicating the safety of voting as directly and clearly to these individuals as possible will yield important results.

Appendix A–Variable descriptions

Table A1. Variable descriptions
Variable Description
COVID-19 risk "Is anyone in your household in a high-risk group for which the annual seasonal influenza vaccine would usually be recommended by the Public Health Agency of Canada? (These conditions include, individuals who are pregnant or those with chronic respiratory disease, chronic heart disease, chronic kidney disease, chronic liver disease, chronic neurological disease, diabetes (all types), cancer, immunosuppression, dysfunction of the spleen, and/or BMI > 40)";

1=Yes
Education No schooling, some elementary, elementary, some secondary, secondary, some technical, technical, some university, bachelor's degree, master's degree, doctorate or professional, rescaled 0-1, don't knows coded as missing
Income No income; 1-30,000; 30,001-60,000; 60,001-90,000; 90,001-110,000; 110,001-150,000; 150,001-200,000; 200+, rescaled 0-1, don't know/prefer not to answer coded as missing
Age In years, rescaled 0-1 from minimum (18) to maximum (99)
Location Rural area, small town, large town, mid-sized city, large city, rescaled 0-1
Female 1=Yes
Region 1=Atlantic; 2=Quebec; 3=Ontario; 4=West
COVID-19 risk perception
  1. "How serious of a threat do you think the coronavirus (COVID-19) is to yourself?"
  2. "How serious of a threat do you think the coronavirus (COVID-19) is to Canadians?"

Not serious at all, not very serious, somewhat serious, very serious, summed and rescaled 0-1

Voted in 2019 1=Yes
In-person voting safety perceptions Very unsafe, somewhat unsafe, somewhat safe, very safe rescaled 0-1
Voting turnout Certain to not vote; unlikely, likely, certain to vote, rescaled 0-1

Appendix B–Estimates for Conjoint Heterogeneous Effects Analysis

Table B1. Estimates for Conjoint Heterogeneous Effects Analysis
Safety Voting
1 2 3 4
Continuous treatment 0.02
(0.02)
-0.00
(0.02)
0.02
(0.02)
-0.01
(0.02)
COVID-19 Risk -0.02
(0.02)
0.01
(0.02)
-0.01
(0.02)
0.02
(0.02)
Risk * Treatment -0.00
(0.01)
-0.00
(0.01)
0.00
(0.01)
-0.00
(0.01)
Age -0.57***
(0.05)
-0.55***
(0.05)
-0.44***
(0.06)
-0.42***
(0.06)
Age * Treatment 0.14***
(0.02)
0.13***
(0.02)
0.11***
(0.02)
0.11***
(0.02)
Education -0.29***
(0.06)
-0.29***
(0.06)
-0.23***
(0.06)
-0.23***
(0.06)
Education * Treatment 0.07***
(0.02)
0.07***
(0.02)
0.07***
(0.02)
0.07***
(0.02)
Location 0.09***
(0.04)
0.12***
(0.03)
0.08**
(0.04)
0.11***
(0.04)
Location * Treatment -0.02*
(0.01)
-0.02**
(0.01)
-0.02
(0.01)
-0.02
(0.01)
Female -0.16***
(0.02)
-0.15***
(0.02)
-0.15***
(0.02)
-0.14***
(0.02)
Female * Treatment 0.03***
(0.01)
0.03***
(0.01)
0.03***
(0.01)
0.03***
(0.01)
COVID-19 risk perception -0.32***
(0.05)
-0.36***
(0.05)
Risk perception * Treatment 0.03*
(0.02)
0.05***
(0.02)
Constant 0.58*** 0.76*** 0.56*** 0.76****
R2 0.18 0.21 0.13 0.16
N 7479 7479 7479 7479

Note: Clustered standard errors in parentheses; * p<0.1, ** p<0.05, *** p<0.01

Appendix C–Word Frequencies

Table C1. Word frequencies, opposition to mandatory masks
Rank 5 character minimum 4 character minimum
Feature Frequency Feature Frequency
1 virus 12 virus 12
2 choic 8 know 11
3 person 8 choic 8
4 right 7 person 8
5 breath 6 sure 8
6 spread 6 right 7
7 mandatori 6 need 7
8 forc 5 breath 6
9 someon 5 spread 6
10 freedom 5 mandatori 6
11 know 5 forc 5
12 distanc 4 want 5
13 go 4 someon 5
14 alreadi 4 freedom 5
15 requir 4 work 5
16 realli 4 stop 5
17 protect 4 distanc 4
18 prevent 4 like 4
19 effect 4 go 4
20 comment 4 alreadi 4

Table C2. Word frequencies, opposition to weekend voting
Rank 5 character minimum 4 character minimum
Feature Frequency Feature Frequency
1 crowd 14 work 39
2 matter 10 time 20
3 disagre 8 crowd 14
4 differ 8 mani 14
5 chang 8 busi 11
6 reason 8 matter 10
7 advanc 7 make 10
8 prefer 7 need 10
9 everyon 6 move 9
10 better 6 sure 9
11 busier 6 disagre 8
12 move 6 differ 8
13 work 6 chang 8
14 religi 5 reason 8
15 relax 5 advanc 7
16 famili 5 prefer 7
17 spread 5 less 7
18 break 4 everyon 6
19 alreadi 4 better 6
20 realli 4 busier 6

Appendix D–Logistic Regression Estimates

Table D1A. Logit estimates, preferred method
Election day Advance polls
1 2 1 2
COVID-19 risk -0.17
(0.14)
-0.13
(0.15)
-0.02
(0.14)
-0.05
(0.14)
Education -0.12
(0.34)
-0.29
(0.35)
-0.19
(0.36)
-0.38
(0.37)
Income 0.00
(0.29)
-0.09
(0.29)
0.32
(0.28)
0.27
(0.29)
Age -0.31
(0.33)
-0.49
(0.34)
1.25***
(0.34)
1.04***
(0.35)
Location 0.22
(0.24)
0.28
(0.24)
0.16
(0.23)
0.16
(0.23)
Female -0.42***
(0.13)
-0.38***
(0.13)
-0.12
(0.14)
-0.08
(0.14)
Quebec 0.14
(0.28)
0.01
(0.29)
0.54*
(0.30)
0.50*
(0.30)
Ontario -0.08
(0.28)
-0.16
(0.28)
0.40
(0.30)
0.36
(0.30)
West -0.08
(0.28)
-0.14
(0.28)
0.35
(0.30)
0.31
(0.31)
COVID-19 risk perception -0.69**
(0.29)
-0.06
(0.28)
Voted in 2019 1.06***
(0.27)
1.28***
(0.30)
Constant -0.53 -0.80* -1.86*** -2.74***
N 1161 1161 1161 1161

Note: Standard errors in parentheses; * p<0.1, ** p<0.05, *** p<0.01

Table D1B. Logit estimates, preferred method
Elections Canada Mail
1 2 1 2
COVID-19 risk 0.56**
(0.17)
0.60**
(0.29)
0.20
(0.15)
0.08
(0.15)
Education -1.34**
(0.63)
-1.37**
(0.66)
1.29***
(0.39)
1.41***
(0.39)
Income 1.21*
(0.65)
1.18*
(0.65)
-0.55*
(0.32)
-0.45
(0.32)
Age -2.79***
(0.77)
-2.82***
(0.76)
0.47
(0.37)
0.62*
(0.37)
Location 0.04
(0.46)
0.10
(0.48)
0.19
(0.26)
0.07
(0.27)
Female -0.35
(0.27)
-0.34
(0.27)
0.38***
(0.15)
0.34**
(0.15)
Quebec 1.10
(1.03)
1.03
(1.02)
-1.15***
(0.32)
-0.93***
(0.34)
Ontario 1.57
(1.00)
1.53
(1.00)
-0.25
(0.29)
-0.12
(0.30)
West 1.53
(1.02)
1.50
(1.02)
-0.28
(0.29)
-0.21
(0.31)
COVID-19 risk perception -0.44
(0.52)
1.62***
(0.34)
Voted in 2019 0.12
(0.40)
-0.72***
(0.22)
Constant -3.22*** -3.00*** -2.03*** -2.70***
N 1161 1161 1161 1161

Note: Standard errors in parentheses; * p<0.1, ** p<0.05, *** p<0.01

Table D2A. Logit estimates, opposed method
Election day Advance polls
1 2 1 2
COVID-19 risk 0.28**
(0.14)
0.23*
(0.14)
0.29*
(0.15)
0.30**
(0.15)
Education -0.10
(0.34)
-0.17
(0.35)
-0.37
(0.37)
-0.41
(0.37)
Income -0.12
(0.30)
-0.22
(0.30)
-0.03
(0.33)
-0.10
(0.33)
Age -0.62**
(0.34)
-0.84**
(0.35)
-1.11***
(0.37)
-1.22***
(0.38)
Location -0.30
(0.22)
-0.36
(0.22)
0.30
(0.24)
0.31
(0.24)
Female -0.09
(0.13)
-0.07
(0.13)
-0.45***
(0.15)
-0.43***
(0.15)
Quebec 0.47
(0.29)
0.54*
(0.30)
0.87**
(0.36)
0.83**
(0.36)
Ontario 0.55*
(0.29)
0.58**
(0.29)
0.59*
(0.35)
0.59*
(0.35)
West 0.34
(0.29)
0.38
(0.29)
0.40
(0.35)
0.41
(0.35)
COVID-19 risk perception 0.50*
(0.29)
-0.25
(0.31)
Voted in 2019 0.49**
(0.20)
0.37
(0.22)
Constant -0.65* -1.23*** -1.17*** -1.24**
N 1124 1124 1124 1124

Note: Standard errors in parentheses; * p<0.1, ** p<0.05, *** p<0.01

Table D2B. Logit estimates, opposed method
Elections Canada Mail
1 2 1 2
COVID-19 risk 0.19
(0.17)
0.19
(0.18)
-0.09
(0.16)
-0.14
(0.16)
Education -0.17
(0.41)
-0.13
(0.42)
0.28
(0.43)
0.30
(0.43)
Income -0.04
(0.37)
0.03
(0.38)
-0.00
(0.34)
0.01
(0.34)
Age -0.95**
(0.44)
-0.83*
(0.45)
0.27
(0.38)
0.27
(0.40)
Location 0.30
(0.28)
0.31
(0.28)
0.17
(0.27)
0.13
(0.27)
Female 0.01
(0.16)
-0.00
(0.16)
0.14
(0.15)
0.13
(0.15)
Quebec 0.33
(0.40)
0.34
(0.41)
0.18
(0.32)
0.27
(0.33)
Ontario 0.70*
(0.39)
0.70*
(0.39)
-0.14
(0.32)
-0.13
(0.32)
West 0.28
(0.40)
0.27
(0.40)
-0.08
(0.32)
-0.07
(0.32)
COVID-19 risk perception 0.04
(0.34)
0.64*
(0.37)
Voted in 2019 -0.32
(0.22)
-0.15
(0.21)
Constant -1.86*** -1.72*** -1.74*** -2.05***
N 1124 1124 1124 1124

Note: Standard errors in parentheses; * p<0.1, ** p<0.05, *** p<0.01

Footnotes

Footnote 1 Minimum and maximum untrimmed weights were 0.66 and 1.66. The average weight is 1 and the standard deviation is 0.14. All results presented below replicate substantively in the absence of weights.