Site MenuAboriginal Electoral Participation in Canada
4. Multivariate Results: The Determinants of Aboriginal Turnout
In order to ascertain the relative and specific impacts attributable to each determinant of voter turnout, we performed logistic regression analyses.Footnote 11 There are two sets of "pooled" models containing multiple surveys. The first includes all variables that were present in all four data sets (2004–2011). As the 2011 study did not include a measure of civic duty, this important factor is absent. Accordingly, we present a second set of analyses with all variables present in the first three studies (2004–2008), including sense of civic duty.
The resulting pooled models cover the largest number of Aboriginal respondents ever analyzed in Canada (over 1,900 respondents in the first set and close to 1,500 in the second). Such samples allow us to draw precise estimates of each variable's contribution to Aboriginal voting behaviour. To draw explicit comparisons with the turnout determinants of non-Aboriginal electors, models for this group are presented alongside.
4.1 Pooled Models
Table 1 presents four sets of results on the determinants of voter turnout among Aboriginal and non-Aboriginal electors. The first two columns deal with the 2004, 2006 and 2008 elections. We then reproduce these results, in the third and fourth columns, with the 2011 election included. This second set of results is required because the 2011 survey did not include questions about sense of civic duty.
Beginning with the results from 2004–2008, the table shows that being registered for the election is a significant predictor of Aboriginal turnout. But far and away, the two driving attitudes among Aboriginal voters are political resources and sense of civic duty. Aboriginal electors who say they are interested in politics, follow politics closely and are familiar with the party platforms are 39 percentage points more likely to vote, even when controlling for all other variables. In the case of civic duty, there is a 27 percentage-point turnout gap between those who feel strongly that voting is a duty and those who do not. The only socio-demographic factor that can compete with these two attitudes is age: older Aboriginals are much more likely to head to the polls on election day than Aboriginal youth.
Two typical but less central socio-demographic predictors also appear: income and education. Aboriginal electors who make less than $20,000 per year are 11 percentage points less likely to vote than those making over $100,000. Individuals with a university education are 7 percentage points more inclined to turn out than Aboriginals who did not graduate from high school. One geographic factor matters as well: lower turnout is associated with residence on a First Nations reserve (a gap of 7 percentage points). The two previous empirical studies of Canadian Aboriginal turnout did not uncover such a finding (Harell et al. 2009, Howe and Bedford 2009).
| Aboriginals (2004–2008) |
Non-Aboriginals (2004–2008) |
Aboriginals (2004–2011) |
Non-Aboriginals (2004–2011) |
|
|---|---|---|---|---|
| Registered | .20** (.02) |
.13** (.01) |
.22** (.02) |
.14** (.01) |
| Living on reserve | -.07** (.02) |
– | -.08** (.02) |
– |
| Women | -.02 (.02) |
-.01 (.01) |
-.03 (.02) |
.01 (.01) |
| Age | .26** (.03) |
.09** (.01) |
.25** (.03) |
.11** (.01) |
| Education | .07** (.03) |
.03* (.01) |
.08** (.03) |
.03** (.01) |
| Income | .11** (.04) |
.02 (.01) |
.09** (.03) |
.04** (.01) |
| Rural residence | -.03 (.02) |
.02** (.01) |
-.02 (.02) |
.02** (.01) |
| Region: British Columbia | .03 (.03) |
.01 (.02) |
.05* (.03) |
-.01 (.01) |
| Region: Ontario | .04 (.03) |
.00 (.01) |
.05* (.03) |
.00 (.01) |
| Region: Quebec | -.06* (.03) |
.01 (.01) |
-.03 (.03) |
.03** (.01) |
| Region: Atlantic | -.05 (.04) |
.01 (.02) |
-.03 (.04) |
.01 (.01) |
| Region: North | .06 (.04) |
.07** (.03) |
.10** (.04) |
.06* (.03) |
| Political resources | .39** (.04) |
.33** (.02) |
.46** (.03) |
.39* *(.01) |
| Civic duty to vote | .27** (.03) |
.26** (.02) |
– | – |
| Parties talk about impor. issues | .03 (.03) |
.03** (.01) |
– | – |
| Saw general EC ad | -.02 (.02) |
.00 (.01) |
-.01 (.02) |
.00 (.01) |
| Saw Aboriginal EC ad | .02 (.02) |
– | .02 (.02) |
– |
| Number of cases | 1473 | 6043 | 1918 | 8596 |
| Pseudo R-squared | .30 | .25 | .28 | .21 |
Cells contain marginal effects of logistic regression coefficients with standard errors in parentheses. All variables range from 0 to 1. **significant at .05; *significant at .10
Many other variables come out as irrelevant in the multivariate analysis. The turnout of men and women is similar. The same can be said for rural and urban residents. Aboriginals from all regions exhibit comparable levels of turnout, though it appears to be slightly lower in Quebec. Believing that political parties address the issues that are personally important to an elector has no effect on the probability of participating. Finally, seeing either a general Elections Canada ad about electoral participation or an ad that specifically targeted Aboriginals to encourage them to vote did not have an impact on Aboriginal turnout. Participation was identical whether people saw one or not.
As can be seen from the second column in Table 1, the determinants of turnout among non-Aboriginal electors are largely the same. Political resources and sense of civic duty are doing most of the heavy lifting, while demographic factors like education are also important. One notable difference, however, is the much greater effect of age on Aboriginal turnout than on non-Aboriginal turnout. Given the noted difficulty in motivating non-Aboriginal youth to vote, this very large effect suggests just how much greater may be the challenge of increasing turnout among Aboriginal youth.
When we consider the models that include the 2011 election (third and fourth columns), the results change little. The effect of political resources increases slightly, but this is likely capturing some of the unmeasured effect from sense of civic duty, as these two concepts are weakly related. Once again, age is pivotal – and starkly more so for Aboriginals than non-Aboriginals. Registration likewise exercises a large effect. Education and income are again relevant. Finally, we note that some regional effects emerge, though they are not substantively remarkable in their size.
To summarize the preceding findings, Chart 1 illustrates the profile of the average Aboriginal and non-Aboriginal voter and non-voter in relation to the key determinants of participation in federal elections.Footnote 12 Focusing first on Aboriginal electors, the chart shows that just half of abstainers recall receiving a voter information card, while this is the case for 87% of Aboriginal voters. Non-voters reside about evenly on and off reserves, but the majority of voters do not live on a reserve. On average, those who vote are 10 years older than those who do not. The typical Aboriginal abstainer has not finished high school, whereas voters have usually obtained their high school diploma. The income of those who turn out is approximately one $20,000-bracket higher than the income of their non-participating counterparts. Just a quarter of Aboriginal non-voters have a level of political resources higher than the scale's midpoint, but close to two thirds of voters surpass that mark. Finally, the belief that voting in elections is a civic duty is shared by 90% of Aboriginals who turn out, much more than the two thirds supporting that idea among Aboriginals who do not exercise their right to vote.
| Non-Aboriginal Non-Voters |
Non-Aboriginal Voters | Aboriginal Non-Voters |
Aboriginal Voters |
|
|---|---|---|---|---|
| Registered | 69% | 93% | 49% | 87% |
| Living on reserve | – | – | 52% | 37% |
| Age | 35–44 | 45–54 | 35–44 | 45–54 |
| Education | College | College | Some high school | Completed high school |
| Income | $40K–$60K | $60K–$80K | $20K–$40K | $40K–$60K |
| Political resources (> midpoint) |
33% | 73% | 24% | 64% |
| Voting is a duty (agree) | 74% | 96% | 68% | 90% |
When we consider the profiles of non-Aboriginal voters and non-voters, the same patterns emerge. Compared to non-Aboriginal abstainers, non-Aboriginal voters are more likely to be registered, older, more educated, wealthier, more concerned by a sense of civic duty, and more interested in, more attentive to and more knowledgeable about politics.Footnote 13 What we see, in sum, is that differences between voters and non-voters among both populations are manifested in each of these key determinants.
A second observation is that when we draw a comparison between Aboriginals and non-Aboriginals in the aggregate (see Table A6 in Appendix 2), the latter group scores higher on every key determinant. On average, non-Aboriginals are more likely to be registered, older, more educated, wealthier, possessing of greater political resources, and slightly more likely to consider voting a duty.
In fact, the gap in turnout between Aboriginal and non-Aboriginal Canadians can be entirely explained by differences in their aggregate profiles. If we simulate the likelihood of Aboriginal turnout when socio-demographic characteristics and attitudinal indicators are equal to those of non-Aboriginals on average (e.g. using Table A6), we find the same predicted rate of electoral participation. This estimate suggests that turnout among Aboriginals would increase by 20 percentage points if their profile on the key determinants matched that of non-Aboriginals, completely closing the gap between them. In short, this turnout gap exists because of substantial differences between the average footing of Aboriginal and non-Aboriginal electors.
4.2 On Reserve vs. Off Reserve
It has been shown that Aboriginals who live on First Nations reserves tend to vote less than those who reside off reserves. But do the two groups have different determinants of turnout? To investigate this possibility, tables A1a and A1b in Appendix 2 present split-sample analyses that consider differences between Aboriginals living on and off reserves. Table A1a presents the results for the 2004–2008 elections, while Table A1b presents results that include the 2011 election.
We focus on the 2004–2008 results, while noting that those including 2011 do not change appreciably. The main picture is one of consistency. The four major determinants of turnout in the analysis of subsection 4.1 – registration, civic duty, political resources and age – have similar effects for both off-reserve and on-reserve Aboriginals, though civic duty appears more important for those living on reserves. That said, there were some slight differences between Aboriginals residing in each of the two settings. While education has a more decisive impact on reserves, income is more critical off reserves. Participation in Quebec diverges according to residence: it is lower than elsewhere on reserves, and it is higher than elsewhere off reserves. Finally, seeing a general Elections Canada ad was only associated with slightly lower turnout on reserves.
4.3 Rural vs. Urban
Could the level of density and economic development be a key cleavage for the structure of Aboriginal turnout determinants? To answer this question, we replicated the models with a second split-sample: whether the respondent's area of residence is urban (third column in tables A1a and A1b) or rural (fourth column in tables A1a and A1b).
Looking at the seven most important determinants of Aboriginal turnout, the nature of the residential area does not play a mediating role. In both rural and urban settings, political resources, civic duty, age, registration, income, education and living on a reserve influence electoral participation in federal elections to an equivalent extent. The only caveat is that the impact of income in rural areas does not attain statistical significance.
Nevertheless, two minor factors do deviate: region and advertising. Turnout was higher in the North, but only in urban localities. Aboriginal electors in rural Ontario and Quebec also behaved differently. Finally, seeing a general Elections Canada ad created a small negative impact, but it was limited to rural settings.
When the 2011 data are included, these effects are essentially unchanged (with the exception that civic duty is not measured).
4.4 Election Differences?
The pooled models of Table 1 could be masking distinct dynamics across the four elections. To verify whether this is the case or not, separate models containing the same variables were conducted for 2004, 2006, 2008 and 2011. These results are presented in Table A2 (Appendix 2).
The four main determinants in the analysis of subsection 4.1 – registration, political resources, civic duty and age – all have similar significant effects in the four models (except for civic duty, which does not appear in 2011). The next trio – living on a reserve, education and income – again show similar impacts, though they often do not reach statistical significance due to the small number of cases in these separate election models.
The Elections Canada ads do not have a significant impact in any of the four elections. Only five of the many variables that did not matter in the pooled models emerge as significant in separate models. Four are regional indicators (British Columbia and Ontario in 2004, and the North in 2006 and 2011), and the other is an attitudinal indicator ("parties talked about important issues" in 2008).
Thus, the patterns observed in the pooled models are essentially consistent across the four elections, and therefore consistent across time.
4.5 The Aboriginal-Specific Explanations
Some may wonder whether our pooled models are losing the Aboriginal aspect of the turnout explanation by retaining only those variables that are common to all four data sets. The separate election models are the perfect place to insert and examine those uncommon variables. In reality, they rarely make a significant contribution above and beyond the central predictors of turnout. These results are demonstrated in Table A3 (Appendix 2).
In 2004, only two of the other variables introduced in bivariate Figure 1 and not included in the pooled models were significant predictors in the multivariate setup: the belief that there should be a guaranteed number of Aboriginals in Parliament and a sense of belonging to Canada. Both tend to suppress Aboriginal turnout. None of the other variables are relevant: neither a sense of belonging to one's Aboriginal group, the desire for more Aboriginal candidates or representation, nor the belief that "my vote does not matter."
In 2006, of the more than 20 extra variables that we can add to the specification, only a handful results in a significant effect. Those who saw the Elections Canada slogan, those who are employed, those who do not have children aged 18 or under, party members, those who do not believe "parties are too influenced by money," those who believe "low turnout weakens democracy", and those who think "more Aboriginal people should vote" were more likely to turn out on election day.Footnote 14 But the top contributors remain age, political resources and sense of civic duty.
In 2008, we can add upwards of 14 other variables. Still, only two matter: Aboriginal respondents tend to participate more when other people in their household voted, and when they do not have a disability.Footnote 15
When we add in the other variables in 2011, the few that have a significant impact are: a sense of connection with one's community, having a disability, employment, and volunteering in one's community.
Thus, these supplementary checks and results clearly indicate that the account of turnout identified in the pooled models is a robust and valid explanation of electoral participation among Aboriginal Canadians.
Return to source of Footnote 11 A logistic model is appropriate when the dependent variable is dichotomous, which is the case for turnout.
Return to source of Footnote 12 For age, education and income, the chart reports the average category among each of the four groups.
Return to source of Footnote 13 There is a gap in the precise education scores of voters and non-voters, even though the summary chart indicates that both groups have the same average education category (college).
Return to source of Footnote 14 The Elections Canada general newspaper ad variable also had a significant positive effect on turnout in 2006, though this was not an additional variable as it was already part of the pooled model (where the medium-specific ad variables of 2006 and 2008 were coded together to match the question format used in 2004 and 2011).
Return to source of Footnote 15 The Elections Canada general television ad variable also had a significant negative effect on turnout in 2008, but it was not a supplementary variable (see the previous footnote).