Inspire Democracy Logo
Site Menu

Voter Turnout among Younger Canadians and Visible Minority Canadians: Evidence from the Provincial Diversity Project

Share this report   Share on Facebook Share on Twitter Share on LinkedIn

6. Opinionation and Voter Turnout

The preceding sections aimed at identifying the characteristics of voters and of habitual non-voters among younger Canadians (those aged 18 to 24 and 25 to 34) and among visible minority Canadians. Among both socio-demographic attributes and political attitudes, many characteristics were identified. In the present section, we offer an exploration into another possibility, namely the role of opinionation or holding opinions (see Krosnick and Milburn, 1990; Delli Carpini and Keeter 1996). Holding an opinion is arguably an important step toward voting. Having an opinion as to whether or not the government cares about the people, whether we can trust public institutions, or even whether it is important to vote is potentially important in explaining why one votes. A certain number of studies examine the characteristics of who hold opinions or provide opinions to questions in political surveys (Krosnick and Milburn 1990; Rapoport 1982; Milbrath and Goel 1977; Francis and Busch 1975). Rare are the studies, however, that examine the consequences of not holding opinions.

Previous research has examined the link between knowledge and voting, but this is not what we are interested in here. Knowledge is important; it is necessary for a proper understanding of the surrounding political environment and for the proper practice of democracy (Delli Carpini and Keeter 1996). Knowledge and opinionation are surely related. It is likely that possessing a greater political knowledge facilitates the development of political opinions; people must "know what something is before they can say how they feel about it, or whether or not they like it" (Cook 1985: 1081). Knowledge and opinionation, thus, might be related, but they are nevertheless different concepts. People who do possess knowledge do not necessarily hold opinions, and conversely, people with weak knowledge can still express firm opinions. Accordingly, the present section aims at exploring whether a link exists between opinionation and the likelihood of voting.

It also appears important to study the link between opinionation and voting for methodological reasons. As reported in the analyses identifying the characteristics of habitual non-voters among younger Canadians in Table 8, only 357 Canadians aged 18 to 24 qualify to be included in the multivariate analyses while 971 are available in the overall PDP sample. Why are not all respondents aged 18 to 24 included in the multivariate analyses? There are two main reasons.

First, given the young age of these respondents, many were simply not eligible to vote in the preceding federal or provincial elections. For instance, about 19% of the 18 to 24 year olds report that they were not eligible to vote in the previous elections, compared to 4% of the 25 to 34 years olds. This reduces the sample size from 971 to 768 respondents. Nothing can be done to include these respondents; we simply cannot investigate whether or not they are habitual non-voters.

Second, and most importantly, Canadians aged 18 to 24 years old are significantly more likely to answer "I don't know" to many questions of the survey, especially those relating to politics. For instance, while about 19% of Canadians aged 18 to 24 years old answer that they don't know how much confidence they have in Elections Canada, this proportion is 11% among the 25 to 34 year olds and only 4% among those aged 35 years and older. Similarly, the proportions of respondents answering that they don't know when asked how much of an impact the federal government has on their lives are 13%, 8% and 3% respectively for the 18-24, 25-34 and 35 and older cohorts.

Table 11 presents the distribution of "don't know" responses to the following political attitude questions that were included in our analyses: feeling close to a political party (federal and provincial), politics is too complicated (federal and provincial), I would feel guilty if I did not vote (federal and provincial), people like me have no say in politics (federal and provincial), I feel my vote can make a difference, attachment to Canada and the respondent's province of residence, the impact of provincial and federal governments on respondent's life, confidence in the House of Commons, in the legislative assembly of the province they reside in, and in Elections Canada. This adds up to a total of sixteen questions.

Table 11. Opinionation among Younger Canadians and Visible Minority Canadians
  Age Groups Visible Minority Status
Number of "don't know" responses (%) 18-24 25-34 35+ Visible minority Canadians Other Canadians
0 57 68 84 72 80
1 7 9 7 6 8
2 9 9 4 7 5
3 9 5 2 7 3
4 3 2 1 2 1
5 6 2 0 2 1
6 and more 8 5 2 5 2
n= 888 2174 4698 5275 1950

Source: Provincial Diversity Project. 

Table 11 indicates that Canadians aged 18 to 24 are less likely than older Canadians to hold a substantive opinion on all sixteen questions. While 84% of Canadians aged 35 and older have provided an opinion in response to all sixteen questions, this proportion drops to 68% among those aged 25 to 34 and to 57% among those aged 18 to 24. In contrast, we observe that while less than 2% of Canadians aged 35 and older answer "I don't know" to at least six of the sixteen questions, this proportion increases to 5% among Canadians aged 25 to 34 and to more than 8% among those aged 18 to 24.Footnote 11

These numbers indicate that many Canadians aged 18 to 24 were not included in the analyses used to identify habitual non-voters because they held no opinion on one or many variables included in the model. The moment that one "don't know" response is provided, the respondent cannot be included in the analyses.Footnote 12 The data presented in Table 11 thus indicate that based on the 16 items examined, we can only keep 57% of Canadians aged 18-24, 68% of those aged 25 to 34 and 84% of those aged 35 years and older for the analyses. The rest of respondents are excluded because they provide one or more "don't know" responses.Footnote 13

Non-response, or a weak level of opinionation, is thus more prevalent among younger Canadians than older Canadians. This finding is consistent with Kronick and Milburn's conclusions (1990). This might not be too surprising given that younger Canadians have accumulated less experience interacting with and observing the Canadian political system than older Canadians have. Sears and Valentino (1997) also observe a lower level of opinination among younger Americans.

Together, the ineligibility to vote and the greater propensity to provide "don't know" responses substantially reduce the sample size available for the analysis of Canadians aged 18 to 24 (from 971 to 357 in the multivariate analysis). This has important implications not just for the sample size available, but also for the characteristics of the sample available for the analyses. As a final step to this report, we propose to examine whether there is a connection between weak opinionation and the propensity to be a habitual non-voter. By doing so, we do not only bring a novel perspective on the reasons why people vote or not, but we also extend the analyses to a group of Canadians that is very often excluded from the analyses investigating the roots of non-voting.

In order to investigate this question, the analyses presented in Table 12 examine, once again, whether or not Canadians are habitual non-voters. This time, however, we limit the investigation to socio-demographic characteristics and include our scale of opinionation that indicates the number of times respondents have selected "I don't know" in response to the sixteen political questions listed above. Because of low frequencies, scores ranging from 6 to 16 are merged together and coded as 6. The scale ranges from 0 to 6. We perform the analyses here for each of the three age categories: 18-24, 25-34, and 35 and older.

Table 12. Opinionation and Habitual Non-Voting among Younger Canadians
  Habitual non-voters (1-0)
  18-24 25-34 35+
Socio-Demographic Factors B SE B SE B SE
Age -.01 .10 .01 .03 -.07 .01 a
Woman -.02. .28 .32 .19 .03 .17
Education (ref. no high school)









Household income -.05 .04 -.16 .03 a -.09 .04 a
Unemployed -.79 .54 .56 .28 c -.91 .43 c
Student .80 .31 c .27 .34 -1.77 1.06
Married -.23 .29 -.05 .18 -.63 .21 b
Time spent at religious inst. -.24 .42 -.84 .29 b -.31 .29
Visible minority .02 .36 .18 .31 .19 .45
Immigrant (ref. Can. Born)
0-10 years
11-20 years
More than 20 years








Regions (ref. Ontario)



Don't know scale (0-6) .31 .09 b .38 .06 a .39 .07 a
Constant .91 2.33 00 .89 3.30 .50 a
Number of observations 488 1657 3727

Source: Provincial Diversity Project. Entries report B coefficients (Binomial logistic regressions)
a: p<.001; b: p<.01; c: p<.05.

First, it is important to highlight the sample available for the analyses. For the age group 18 to 24, our sample size is 488 as opposed to the 357 that would have been eligible if all political variables were included in the model. For the 25 to 34 year old group, the sample size is 1657 (as opposed to 1309), and for the 35 and older group, the sample size is 3727 (as opposed to 3355). The sample sizes for the 18 to 24 and 25 to 34 age groups thus increase by 37% and 27% respectively, while the sample size of those aged 35 years and older increases by only about 11%. These sample-size increases are important. They highlight the extent to which, among younger voters, we lose a substantial proportion of the sample if we exclude them on the basis of "don't know" responses.Footnote 14

Second, the findings further support the importance of "don't know" responses by showing that the scale of opinionation (or number of "don't know" responses) is significantly related to the propensity of being a habitual non-voter. The larger the number of "don't know" responses, the greater the likelihood of being a habitual non-voter. This holds among all three age groups. Figure 1 reports the predicted probabilities of being a habitual non-voter by level of "don't know" responses for each of the three age groups. The results clearly highlight the increasing probability of being a habitual non-voter among all three age groups as the number of "don't know" responses increases. Moreover, it is striking to see that the lower propensity to vote is not limited to those who provide a large number of "don't know" responses. The relationship appears linear; even a single mention of a "don't know" response increases the probability of being a habitual non-voter. Any indication of low opinionation, no matter how minor, increases the likelihood of being a habitual non-voter.Footnote 15

Figure 1: Predicted Probability of Being a Habitual Non...
Text version of "Figure 1: Predicted Probability of Being a Habitual Non.."

Is the propensity to answer "don't know" simply the reflection of a weak interest in politics?  Interest in politics negatively correlates with the propensity to provide "don't know" responses; the greater the interest in politics, the smaller the number of "don't know" responses. The correlation coefficients are -.37, -.34, and -.27 respectively for Canadians aged 18-24, 25-34 and 35 and older. We further investigated the link between "don't know" responses and interest in politics by adding the latter variable in our multivariate analyses. We do not present the full table here but Figure 2 reports the predicted probabilities of being a habitual non-voter according to the number of "don't know" responses provided by respondents once controlling for interest in politics. In contrast to Figure 1, the slopes in Figure 2 are not as steep. Including interest in politics thus accounts for a portion of the link between "don't know" responses and habitual non-voting. One of the reasons why the people who report "don't know" responses do not vote is that they are not interested in politics. But this is not the entire story; the relationship between opinionation and habitual non-voting remains significant even when controlling for interest in politics. And the effect is substantial; if we performed a simulation like the ones we performed in the previous section, setting the mean level of opinionation of younger Canadians (1.44) to that observed for older Canadians (.36), the model predicts that the level of habitual non-voting would be about 5 points lower among younger Canadians. Of course, part of the effect would likely be explained by other political attitudes presently not explained, but this nevertheless signals a non-negligible importance of the opinionation variable.

Figure 2: Predicted Probability of Being a Habitual Non...
Text version of "Figure 2: Predicted Probability of Being a Habitual Non.."

This suggests that one of the reasons that Canadians do not vote is a low level of opinionation. A significant proportion of Canadians do not appear to have developed firm opinions about their attachment to the political system, their sense of civic duty, or their evaluation of government responsiveness. For them, voting appears to be a less appealing activity. The relationship holds among all three age groups, but as demonstrated in Table 10, younger Canadians exhibit substantially lower levels of opinionation than older Canadians. It is worth emphasizing that our analyses do not control for respondents' knowledge of Canadian politics. Unfortunately, such indicators are not available in the Provincial Diversity Project. It is possible that the relationship between opinionation and habitual non-voting could be partially or fully accounted for if indicators of respondents' knowledge of politics were included in the model. Delli Carpini and Keeter argue that knowledge is even the strongest predictor of opinionation in their analyses (1996:230).

Footnote 11 Whether the reality of high "don't know" response rates in the Provincial Diversity Project among the youth is something unique to this data is a possibility; investigating whether this is also the case in other political surveys like the Canadians Election Study is a task beyond the scope of this current report. The difference in the propensity for "don't know" responses between visible minorities and other Canadians is more limited. Accordingly, we limit this section of the report to the analyses of younger Canadians.

Footnote 12 Because listwise deletion method is used.

Footnote 13 Other methods can be used when conducting multivariate analyses that allow including respondents who provide non-responses such a imputing sample means for missing cases. We prefer not to use such methods. The following analyses indicate that respondents who provide don't know as a response may have a different political profile and propensity to vote, and accordingly we should treat their missing information as such instead of trying to impute a response that they did not provide. We acknowledge, however, that such a methodological decision is open for debate in the field and ultimately further analyses should be performed to better understand both the meaning of these non-responses and what kinds of implications using different methods to treat the missing information have on the results of our inferences.

Footnote 14 There are still a large numbers of missing cases because of the household income question.

Footnote 15 Other specifications were tested to verify that the relationship was indeed linear. Among visible minority Canadians, we also observe that a greater propensity to provide "don't know" responses is associated with a greater likelihood to be a habitual non-voter (results not presented).