Generational Change: Looking at Declining Youth Voter Turnout over Time
Peter Loewen, PhD
University of Toronto/Loewen FRM Inc
November 30, 2013
Background
In 2008, André Blais and I conducted a study on youth voter turnout commissioned by Elections Canada. The study, Youth Electoral Engagement in Canada, was subsequently updated in January 2011 prior to the May 2011 general federal election. The study drew upon data from the Canadian Election Studies (CES) to estimate voter turnout by age cohort for every election between 1965 and 2008 (with the exception of 1972).
This research update extends the analysis to include the May 2011 election. It is divided into two parts. First, a recalculated analysis of estimated voter turnout by cohort and election is presented. Second, four graphics of key findings are presented that can be used to visually convey the patterns that emerge from this analysis. These graphics can be used by researchers and other stakeholders to illustrate declining youth voter turnout in Canada and its associated trends.
1. Updated cohort and election turnout analysis
Table 1 presents updated estimates of turnout by cohort and election. Cohorts are defined by the first election in which a voter was eligible to participate. Each cell presents the probability of an individual voting given their cohort and the election. As in previous iterations, the results are derived from a logistic regression in which voter turnout is modeled as a function of the first election in which a voter was eligible (their so called "cohort"), the election, a voter's age, and a squared-transformation of a voter's age.Footnote 1 The results are weighted according to observed actual turnout in each election. Full regression results are presented in Table 2.
1965 | 1968 | 1974 | 1979 | 1980 | 1984 | 1988 | 1993 | 1997 | 2000 | 2004 | 2006 | 2008 | 2011 | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Cohort | ||||||||||||||
1965 | 66 | 69 | 68 | 77 | 70 | 79 | 82 | 83 | 78 | 77 | 75 | 77 | 73 | 73 |
1968 | 71 | 71 | 79 | 73 | 81 | 84 | 85 | 81 | 80 | 78 | 80 | 77 | 78 | |
1972 | 61 | 71 | 64 | 73 | 77 | 80 | 74 | 73 | 71 | 73 | 71 | 72 | ||
1974 | 56 | 67 | 59 | 70 | 74 | 77 | 72 | 70 | 69 | 71 | 69 | 69 | ||
1979 | 61 | 53 | 64 | 69 | 73 | 67 | 66 | 64 | 67 | 64 | 65 | |||
1980 | 49 | 61 | 67 | 71 | 64 | 63 | 62 | 65 | 62 | 63 | ||||
1984 | 57 | 62 | 67 | 60 | 59 | 58 | 61 | 58 | 60 | |||||
1988 | 54 | 59 | 52 | 52 | 51 | 54 | 51 | 53 | ||||||
1993 | 54 | 48 | 47 | 46 | 50 | 47 | 49 | |||||||
1997 | 42 | 41 | 41 | 44 | 42 | 44 | ||||||||
2000 | 35 | 35 | 38 | 36 | 38 | |||||||||
2004 | 34 | 38 | 36 | 38 | ||||||||||
2006 | 46 | 43 | 46 | |||||||||||
2008 | 27 | 29 | ||||||||||||
2011 | 35 |
This table presents estimates of the rate of voter turnout for each cohort in each federal election between 1965 and 2011, except 1972. Cohorts are defined by the first election in which a voter was eligible to participate.
Coef. | S.E. | z | P>z | |
---|---|---|---|---|
Age | 0.03 | 0.00 | 6.95 | 0.00 |
Age-squared | 0.00 | 0.00 | -7.43 | 0.00 |
1965 cohort | -0.20 | 0.06 | -3.24 | 0.00 |
1968 cohort | -0.05 | 0.07 | -0.71 | 0.48 |
1972 cohort | -0.41 | 0.06 | -7.15 | 0.00 |
1974 cohort | -0.51 | 0.07 | -6.89 | 0.00 |
1979 cohort | -0.70 | 0.07 | -10.27 | 0.00 |
1980 cohort | -0.78 | 0.10 | -7.49 | 0.00 |
1984 cohort | -0.92 | 0.08 | -11.49 | 0.00 |
1988 cohort | -1.18 | 0.09 | -13.30 | 0.00 |
1993 cohort | -1.28 | 0.10 | -13.34 | 0.00 |
1997 cohort | -1.44 | 0.11 | -13.10 | 0.00 |
2000 cohort | -1.64 | 0.13 | -12.69 | 0.00 |
2004 cohort | -1.58 | 0.14 | -11.32 | 0.00 |
2006 cohort | -1.20 | 0.23 | -5.31 | 0.00 |
2008 cohort | -1.91 | 0.25 | -7.65 | 0.00 |
2011 cohort | -1.57 | 0.30 | -5.24 | 0.00 |
1965 election | 0.28 | 0.14 | 1.95 | 0.05 |
1968 election | 0.33 | 0.14 | 2.33 | 0.02 |
1974 election | 0.21 | 0.15 | 1.48 | 0.14 |
1979 election | 0.57 | 0.15 | 3.80 | 0.00 |
1980 election | 0.24 | 0.15 | 1.54 | 0.12 |
1984 election | 0.63 | 0.15 | 4.10 | 0.00 |
1988 election | 0.79 | 0.16 | 4.92 | 0.00 |
1993 election | 0.89 | 0.17 | 5.32 | 0.00 |
1997 election | 0.55 | 0.17 | 3.18 | 0.00 |
2000 election | 0.47 | 0.18 | 2.64 | 0.01 |
2004 election | 0.38 | 0.19 | 2.04 | 0.04 |
2006 election | 0.49 | 0.19 | 2.59 | 0.01 |
2008 election | 0.36 | 0.20 | 1.83 | 0.07 |
2011 election | 0.40 | 0.20 | 2.01 | 0.04 |
N | 40,387 |
---|---|
Log likelihood | -23,548 |
Prob >chi2 | 0.00 |
This table presents the econometric estimates supporting Table 1. Data are drawn from the 1965, 1968, 1974, 1979, 1984, 1988, 1993, 1997, 2000, 2004, 2006, 2008, and 2011 Canadian Election Studies. The dependent variable is voted (1) or did not vote (0) in the election. The model is a logistic regression. The cohort variables indicate the election in which a respondent was first eligible to vote. Accordingly, those coefficients report the effect of membership in a certain cohort on overall turnout. The election variables indicate the effect of a respective election, given a cohort.
2. Graphical presentations
In this section, four graphical renderings of the above data are presented. The goal of these graphs is to present cleanly and simply the key findings in Table 1.
Figure 1. Voter turnout by first time electors, 1965-2011
(For text description of graphic, see next paragraph.)
This graph demonstrates estimated voter turnout by electors eligible to vote for the first time in every election between 1965-2011 (except 1972). The graph demonstrates a strong negative trend in first election turnout in the last 50 years. First-time turnout reached a high of 71% in 1968. The average in the last four elections is 36%. This graph is useful for demonstrating the decline in first-election participation in more recent cohorts.
Figure 2. Average voter turnout by cohorts, 1965-2011
(For text description of graphic, see next paragraph.)
This graph demonstrates the average estimated voter turnout for cohorts grouped together by decade for every election between 1965-2011 (except 1972). For example, the '1960s' cohort represents all those cohorts who first became eligible to vote during that decade. Average turnout is calculated by taking the simple average of the estimated turnout for each cohort within a decade. The blue line at the top represents the 1960s cohort, the red line below that the 1970s cohort, the green line below that the 1980s cohort, the orange line below that the 1990s cohort, and the light blue line at the bottom, the cohort from the 2000s.
The graph demonstrates two things. First, all applicable cohorts experienced increasing voter turnout through the 1993 election. At that time, average turnout began declining. Second, the most important differences are not changes in over-time turnout within cohorts. Instead, the most important differences are found in the lower starting point of each successive cohort. While the 1960s cohort started at a turnout rate of approximately 70%, it is only half as large among the 2000s cohort. This graph is most useful for demonstrating the lower levels of turnout by successive cohorts.
Figure 3. Turnout gaps between cohorts
(For text description of graphic, see next paragraph.)
This figure demonstrates the turnout gap between those cohorts with the highest levels of turnout and those with the lowest in any given election. At the start of the time series, the difference in participation rates is 24 percentage points. By 2011, the gap climbs to 49 percentage points. This suggests an increasing generational inequality in electoral participation. The series is limited to elections from 1980 on to allow for a sufficiently large number of comparison groups, as there are no data for older cohorts in the earlier elections in our sample. This graph is most useful for showing the increasing gap in turnout between cohorts.
Figure 4. Maximum and minimum cohort turnout by election
(For text description of graphic, see next paragraph.)
This graph presents the data used to generate Figure 3. The blue line reports the voter turnout rate for the cohort with the highest level of turnout in an election. The red reports the turnout for the cohort with the lowest level. The most important observation from these figures is that the widening gap in turnout is not attributable to an increasing rate of participation among the most participatory cohorts. The turnout of this group ranges only 8 percentage points since 1980. By contrast, the turnout of the least participatory cohort has declined substantially, from a high of 57 percentage points in 1984 to 27 points in 2008. The widening gap in turnout then is due to increasing abstention at the bottom end, rather than increasing participation at the top end. This graph is most useful for demonstrating the source of the turnout gap between cohorts.
Return to source of Footnote 1 The results presented in Table 1 are highly reflective of previous estimates. Two factors are of note, however. First, the estimates in any given cell are likely to differ slightly from earlier estimates. This is as a result of the additional data added by the 2011 election and the resulting econometric re-estimation. Second, the estimates for later cohorts, in particular 2006 and 2008, appear higher than previous estimates.