Skip to contents

Are Quebec elections decided during the campaign?

Source: vignettes/articles/transitions.Rmd
transitions.Rmd

Version française

This page downloads 5 studies (about 5 MB) the first time it runs. options(qesR.cache = "disk") keeps them on disk for later sessions.

Since the 1988 federal election, Canadian election studies have shown that opinion can move a great deal between the start of a campaign and the vote (Johnston et al. 1992). It is often said that Quebec elections are decided at the last minute, a claim that returns whenever the polls miss the result, as they did in 1998 and 2018. In 1998 the polls had the Parti Québécois (PQ) ahead, yet the Parti libéral du Québec (PLQ) took more of the valid votes (43.6% compared to 42.9%). In 2018 they forecast a close race between the PLQ and the Coalition avenir Québec (CAQ), which won by 13 points. Each time, part of the explanation was sought among voters who changed their minds at the end of the campaign and among non-disclosers, who would not reveal their intention. Durand and Blais (1999) reject both a late Liberal rise and a lower PQ turnout in 1998, and trace the error to sampling and to how non-respondents behaved; for 2018, they conclude that last-minute shifts and the vote of non-disclosers explain most of the gap (Durand and Blais 2020).

However, a cross-section tells us how the vote ended up, not who moved. Panels can, because they interview the same people during the campaign and after the vote. Five Quebec election surveys did so, from the 1998 CROP and CREATEC polls to the 2022 Quebec Election Study, including the panels Claire Durand ran in 2007, 2012 and 2018. We use them to ask a simple question: how many Quebec voters change their minds during a campaign, and who gains from it? Between 72% and 78% of those who voted reported the party they had named during the campaign. The undecided were a small group, from 6% to 10% of voters, and voters who switched from one party to another outnumbered them in every panel except 2018, where the two groups were about the same size. On net, these moves rarely changed a party’s share by more than a few points. The largest gain is the one the 2018 post-mortem pointed to: among the same respondents, the CAQ gained 6.1 points between intention and vote, and 37% of that year’s undecided voted CAQ. In short, a Quebec campaign moves many voters but, on net, few votes.

Most voters did what they said

Most voters did what they said; in 2018 the CAQ drew from every side, including 12% of those who intended to vote PQFrom the intention during the campaign (rows) to the vote reported after the election (columns): the percentage of each row
Two heatmaps (2018 panel, QES 2022): rows are the vote intention during the campaign, with the number of respondents, columns the vote reported after the election, each cell the percentage of the row. The diagonal dominates: in 2022, 83% of those who intended to vote CAQ reported a CAQ vote and 67% of those who intended to vote QS a QS vote; 24% of the undecided did not vote. Values in the table view.Two heatmaps (2018 panel, QES 2022): rows are the vote intention during the campaign, with the number of respondents, columns the vote reported after the election, each cell the percentage of the row. The diagonal dominates: in 2022, 83% of those who intended to vote CAQ reported a CAQ vote and 67% of those who intended to vote QS a QS vote; 24% of the undecided did not vote. Values in the table view.

Source: qesR. Respondents interviewed during the campaign and after the election, in the 2018 Durand panel and the QES 2022, weighted with the post-election weight. Rows: the party intended, with the number of respondents; the undecided named no party, even when asked (where the panel asked) which party they leaned to. Each row includes those who did not vote. Grey cells have no value: the 2018 panel did not list the PCQ (n.l.), the 2022 campaign question had no 'would not vote' answer, and rows of fewer than 30 respondents are not drawn. Confidence intervals and the 1998, 2007 and 2012 panels are in the table view.

Table view
Election Study Intention Reported vote Row %, [95% CI] n (row) Weighting
1998 1998 polls PLQ PLQ 80.1 [75.3, 84.1] 311 unweighted: no validated weight
1998 1998 polls PLQ PQ 5.1 [3.2, 8.2] 311 unweighted: no validated weight
1998 1998 polls PLQ ADQ 4.5 [2.7, 7.5] 311 unweighted: no validated weight
1998 1998 polls PLQ QS — 311 unweighted: no validated weight
1998 1998 polls PLQ CAQ — 311 unweighted: no validated weight
1998 1998 polls PLQ PCQ — 311 unweighted: no validated weight
1998 1998 polls PLQ Other 0.3 [0.0, 2.2] 311 unweighted: no validated weight
1998 1998 polls PLQ Did not vote 10.0 [7.1, 13.8] 311 unweighted: no validated weight
1998 1998 polls PQ PLQ 3.9 [2.5, 6.2] 433 unweighted: no validated weight
1998 1998 polls PQ PQ 85.7 [82.1, 88.7] 433 unweighted: no validated weight
1998 1998 polls PQ ADQ 2.5 [1.4, 4.5] 433 unweighted: no validated weight
1998 1998 polls PQ QS — 433 unweighted: no validated weight
1998 1998 polls PQ CAQ — 433 unweighted: no validated weight
1998 1998 polls PQ PCQ — 433 unweighted: no validated weight
1998 1998 polls PQ Other 0.9 [0.3, 2.4] 433 unweighted: no validated weight
1998 1998 polls PQ Did not vote 6.9 [4.9, 9.7] 433 unweighted: no validated weight
1998 1998 polls ADQ PLQ 13.1 [9.5, 17.8] 259 unweighted: no validated weight
1998 1998 polls ADQ PQ 15.4 [11.5, 20.4] 259 unweighted: no validated weight
1998 1998 polls ADQ ADQ 56.0 [49.9, 61.9] 259 unweighted: no validated weight
1998 1998 polls ADQ QS — 259 unweighted: no validated weight
1998 1998 polls ADQ CAQ — 259 unweighted: no validated weight
1998 1998 polls ADQ PCQ — 259 unweighted: no validated weight
1998 1998 polls ADQ Other 1.2 [0.4, 3.5] 259 unweighted: no validated weight
1998 1998 polls ADQ Did not vote 14.3 [10.5, 19.1] 259 unweighted: no validated weight
1998 1998 polls Other PLQ n < 30 26 unweighted: no validated weight
1998 1998 polls Other PQ n < 30 26 unweighted: no validated weight
1998 1998 polls Other ADQ n < 30 26 unweighted: no validated weight
1998 1998 polls Other QS — 26 unweighted: no validated weight
1998 1998 polls Other CAQ — 26 unweighted: no validated weight
1998 1998 polls Other PCQ — 26 unweighted: no validated weight
1998 1998 polls Other Other n < 30 26 unweighted: no validated weight
1998 1998 polls Other Did not vote n < 30 26 unweighted: no validated weight
1998 1998 polls None / would not vote PLQ n < 30 22 unweighted: no validated weight
1998 1998 polls None / would not vote PQ n < 30 22 unweighted: no validated weight
1998 1998 polls None / would not vote ADQ n < 30 22 unweighted: no validated weight
1998 1998 polls None / would not vote QS — 22 unweighted: no validated weight
1998 1998 polls None / would not vote CAQ — 22 unweighted: no validated weight
1998 1998 polls None / would not vote PCQ — 22 unweighted: no validated weight
1998 1998 polls None / would not vote Other n < 30 22 unweighted: no validated weight
1998 1998 polls None / would not vote Did not vote n < 30 22 unweighted: no validated weight
1998 1998 polls Undecided PLQ 38.3 [28.4, 49.3] 81 unweighted: no validated weight
1998 1998 polls Undecided PQ 28.4 [19.7, 39.1] 81 unweighted: no validated weight
1998 1998 polls Undecided ADQ 13.6 [7.7, 22.9] 81 unweighted: no validated weight
1998 1998 polls Undecided QS — 81 unweighted: no validated weight
1998 1998 polls Undecided CAQ — 81 unweighted: no validated weight
1998 1998 polls Undecided PCQ — 81 unweighted: no validated weight
1998 1998 polls Undecided Other 2.5 [0.6, 9.3] 81 unweighted: no validated weight
1998 1998 polls Undecided Did not vote 17.3 [10.5, 27.1] 81 unweighted: no validated weight
2007 2007 panel PLQ PLQ 73.2 [68.6, 77.3] 395 unweighted: no validated weight
2007 2007 panel PLQ PQ 3.0 [1.7, 5.3] 395 unweighted: no validated weight
2007 2007 panel PLQ ADQ 8.1 [5.8, 11.2] 395 unweighted: no validated weight
2007 2007 panel PLQ QS 0.5 [0.1, 2.0] 395 unweighted: no validated weight
2007 2007 panel PLQ CAQ — 395 unweighted: no validated weight
2007 2007 panel PLQ PCQ — 395 unweighted: no validated weight
2007 2007 panel PLQ Other 0.8 [0.2, 2.3] 395 unweighted: no validated weight
2007 2007 panel PLQ Did not vote 14.4 [11.3, 18.3] 395 unweighted: no validated weight
2007 2007 panel PQ PLQ 3.8 [2.3, 6.1] 399 unweighted: no validated weight
2007 2007 panel PQ PQ 67.4 [62.7, 71.8] 399 unweighted: no validated weight
2007 2007 panel PQ ADQ 9.0 [6.6, 12.3] 399 unweighted: no validated weight
2007 2007 panel PQ QS 1.5 [0.7, 3.3] 399 unweighted: no validated weight
2007 2007 panel PQ CAQ — 399 unweighted: no validated weight
2007 2007 panel PQ PCQ — 399 unweighted: no validated weight
2007 2007 panel PQ Other 1.8 [0.8, 3.6] 399 unweighted: no validated weight
2007 2007 panel PQ Did not vote 16.5 [13.2, 20.5] 399 unweighted: no validated weight
2007 2007 panel ADQ PLQ 4.3 [2.7, 6.9] 370 unweighted: no validated weight
2007 2007 panel ADQ PQ 11.1 [8.3, 14.7] 370 unweighted: no validated weight
2007 2007 panel ADQ ADQ 71.9 [67.1, 76.2] 370 unweighted: no validated weight
2007 2007 panel ADQ QS 1.1 [0.4, 2.8] 370 unweighted: no validated weight
2007 2007 panel ADQ CAQ — 370 unweighted: no validated weight
2007 2007 panel ADQ PCQ — 370 unweighted: no validated weight
2007 2007 panel ADQ Other 1.4 [0.6, 3.2] 370 unweighted: no validated weight
2007 2007 panel ADQ Did not vote 10.3 [7.6, 13.8] 370 unweighted: no validated weight
2007 2007 panel QS PLQ 5.1 [1.9, 12.7] 79 unweighted: no validated weight
2007 2007 panel QS PQ 21.5 [13.8, 31.9] 79 unweighted: no validated weight
2007 2007 panel QS ADQ 13.9 [7.9, 23.4] 79 unweighted: no validated weight
2007 2007 panel QS QS 41.8 [31.4, 52.9] 79 unweighted: no validated weight
2007 2007 panel QS CAQ — 79 unweighted: no validated weight
2007 2007 panel QS PCQ — 79 unweighted: no validated weight
2007 2007 panel QS Other 10.1 [5.1, 19.0] 79 unweighted: no validated weight
2007 2007 panel QS Did not vote 7.6 [3.5, 15.9] 79 unweighted: no validated weight
2007 2007 panel Other PLQ 13.9 [8.4, 22.1] 101 unweighted: no validated weight
2007 2007 panel Other PQ 15.8 [9.9, 24.3] 101 unweighted: no validated weight
2007 2007 panel Other ADQ 12.9 [7.6, 20.9] 101 unweighted: no validated weight
2007 2007 panel Other QS 3.0 [1.0, 8.8] 101 unweighted: no validated weight
2007 2007 panel Other CAQ — 101 unweighted: no validated weight
2007 2007 panel Other PCQ — 101 unweighted: no validated weight
2007 2007 panel Other Other 33.7 [25.1, 43.4] 101 unweighted: no validated weight
2007 2007 panel Other Did not vote 20.8 [14.0, 29.8] 101 unweighted: no validated weight
2007 2007 panel None / would not vote PLQ n < 30 29 unweighted: no validated weight
2007 2007 panel None / would not vote PQ n < 30 29 unweighted: no validated weight
2007 2007 panel None / would not vote ADQ n < 30 29 unweighted: no validated weight
2007 2007 panel None / would not vote QS n < 30 29 unweighted: no validated weight
2007 2007 panel None / would not vote CAQ — 29 unweighted: no validated weight
2007 2007 panel None / would not vote PCQ — 29 unweighted: no validated weight
2007 2007 panel None / would not vote Other n < 30 29 unweighted: no validated weight
2007 2007 panel None / would not vote Did not vote n < 30 29 unweighted: no validated weight
2007 2007 panel Undecided PLQ 19.2 [12.8, 27.9] 104 unweighted: no validated weight
2007 2007 panel Undecided PQ 26.9 [19.3, 36.2] 104 unweighted: no validated weight
2007 2007 panel Undecided ADQ 23.1 [16.0, 32.1] 104 unweighted: no validated weight
2007 2007 panel Undecided QS 0.0 104 unweighted: no validated weight
2007 2007 panel Undecided CAQ — 104 unweighted: no validated weight
2007 2007 panel Undecided PCQ — 104 unweighted: no validated weight
2007 2007 panel Undecided Other 4.8 [2.0, 11.0] 104 unweighted: no validated weight
2007 2007 panel Undecided Did not vote 26.0 [18.4, 35.2] 104 unweighted: no validated weight
2012 2012 panel PLQ PLQ 82.3 [75.5, 87.5] 158 unweighted: no validated weight
2012 2012 panel PLQ PQ 3.2 [1.3, 7.4] 158 unweighted: no validated weight
2012 2012 panel PLQ ADQ — 158 unweighted: no validated weight
2012 2012 panel PLQ QS 1.3 [0.3, 4.9] 158 unweighted: no validated weight
2012 2012 panel PLQ CAQ 6.3 [3.4, 11.4] 158 unweighted: no validated weight
2012 2012 panel PLQ PCQ n.l. 158 unweighted: no validated weight
2012 2012 panel PLQ Other 0.6 [0.1, 4.4] 158 unweighted: no validated weight
2012 2012 panel PLQ Did not vote 6.3 [3.4, 11.4] 158 unweighted: no validated weight
2012 2012 panel PQ PLQ 1.0 [0.2, 3.7] 210 unweighted: no validated weight
2012 2012 panel PQ PQ 89.0 [84.1, 92.6] 210 unweighted: no validated weight
2012 2012 panel PQ ADQ — 210 unweighted: no validated weight
2012 2012 panel PQ QS 2.4 [1.0, 5.6] 210 unweighted: no validated weight
2012 2012 panel PQ CAQ 1.0 [0.2, 3.7] 210 unweighted: no validated weight
2012 2012 panel PQ PCQ n.l. 210 unweighted: no validated weight
2012 2012 panel PQ Other 0.0 210 unweighted: no validated weight
2012 2012 panel PQ Did not vote 6.7 [4.0, 10.9] 210 unweighted: no validated weight
2012 2012 panel QS PLQ 0.0 47 unweighted: no validated weight
2012 2012 panel QS PQ 27.7 [16.8, 42.0] 47 unweighted: no validated weight
2012 2012 panel QS ADQ — 47 unweighted: no validated weight
2012 2012 panel QS QS 57.4 [43.1, 70.7] 47 unweighted: no validated weight
2012 2012 panel QS CAQ 4.3 [1.1, 15.5] 47 unweighted: no validated weight
2012 2012 panel QS PCQ n.l. 47 unweighted: no validated weight
2012 2012 panel QS Other 4.3 [1.1, 15.5] 47 unweighted: no validated weight
2012 2012 panel QS Did not vote 6.4 [2.1, 18.0] 47 unweighted: no validated weight
2012 2012 panel CAQ PLQ 6.6 [3.7, 11.5] 167 unweighted: no validated weight
2012 2012 panel CAQ PQ 9.0 [5.5, 14.4] 167 unweighted: no validated weight
2012 2012 panel CAQ ADQ — 167 unweighted: no validated weight
2012 2012 panel CAQ QS 2.4 [0.9, 6.2] 167 unweighted: no validated weight
2012 2012 panel CAQ CAQ 72.5 [65.2, 78.7] 167 unweighted: no validated weight
2012 2012 panel CAQ PCQ n.l. 167 unweighted: no validated weight
2012 2012 panel CAQ Other 1.2 [0.3, 4.7] 167 unweighted: no validated weight
2012 2012 panel CAQ Did not vote 8.4 [5.0, 13.7] 167 unweighted: no validated weight
2012 2012 panel Other PLQ 10.5 [4.0, 24.9] 38 unweighted: no validated weight
2012 2012 panel Other PQ 18.4 [9.0, 33.9] 38 unweighted: no validated weight
2012 2012 panel Other ADQ — 38 unweighted: no validated weight
2012 2012 panel Other QS 10.5 [4.0, 24.9] 38 unweighted: no validated weight
2012 2012 panel Other CAQ 5.3 [1.3, 18.8] 38 unweighted: no validated weight
2012 2012 panel Other PCQ n.l. 38 unweighted: no validated weight
2012 2012 panel Other Other 50.0 [34.6, 65.4] 38 unweighted: no validated weight
2012 2012 panel Other Did not vote 5.3 [1.3, 18.8] 38 unweighted: no validated weight
2012 2012 panel None / would not vote PLQ n < 30 8 unweighted: no validated weight
2012 2012 panel None / would not vote PQ n < 30 8 unweighted: no validated weight
2012 2012 panel None / would not vote ADQ — 8 unweighted: no validated weight
2012 2012 panel None / would not vote QS n < 30 8 unweighted: no validated weight
2012 2012 panel None / would not vote CAQ n < 30 8 unweighted: no validated weight
2012 2012 panel None / would not vote PCQ n.l. 8 unweighted: no validated weight
2012 2012 panel None / would not vote Other n < 30 8 unweighted: no validated weight
2012 2012 panel None / would not vote Did not vote n < 30 8 unweighted: no validated weight
2012 2012 panel Undecided PLQ 29.5 [19.4, 42.1] 61 unweighted: no validated weight
2012 2012 panel Undecided PQ 23.0 [14.1, 35.1] 61 unweighted: no validated weight
2012 2012 panel Undecided ADQ — 61 unweighted: no validated weight
2012 2012 panel Undecided QS 3.3 [0.8, 12.2] 61 unweighted: no validated weight
2012 2012 panel Undecided CAQ 14.8 [7.9, 26.0] 61 unweighted: no validated weight
2012 2012 panel Undecided PCQ n.l. 61 unweighted: no validated weight
2012 2012 panel Undecided Other 4.9 [1.6, 14.2] 61 unweighted: no validated weight
2012 2012 panel Undecided Did not vote 24.6 [15.4, 36.9] 61 unweighted: no validated weight
2018 2018 panel PLQ PLQ 79.3 [72.6, 84.7] 226 weighted
2018 2018 panel PLQ PQ 0.0 226 weighted
2018 2018 panel PLQ ADQ — 226 weighted
2018 2018 panel PLQ QS 0.8 [0.2, 3.6] 226 weighted
2018 2018 panel PLQ CAQ 5.8 [3.1, 10.7] 226 weighted
2018 2018 panel PLQ PCQ n.l. 226 weighted
2018 2018 panel PLQ Other 2.0 [0.9, 4.6] 226 weighted
2018 2018 panel PLQ Did not vote 12.1 [8.0, 17.9] 226 weighted
2018 2018 panel PQ PLQ 0.7 [0.2, 2.8] 115 weighted
2018 2018 panel PQ PQ 69.3 [56.0, 80.0] 115 weighted
2018 2018 panel PQ ADQ — 115 weighted
2018 2018 panel PQ QS 0.0 115 weighted
2018 2018 panel PQ CAQ 11.7 [5.5, 23.1] 115 weighted
2018 2018 panel PQ PCQ n.l. 115 weighted
2018 2018 panel PQ Other 0.0 115 weighted
2018 2018 panel PQ Did not vote 18.3 [9.6, 32.0] 115 weighted
2018 2018 panel QS PLQ 0.0 109 weighted
2018 2018 panel QS PQ 9.6 [5.1, 17.6] 109 weighted
2018 2018 panel QS ADQ — 109 weighted
2018 2018 panel QS QS 67.8 [56.9, 77.1] 109 weighted
2018 2018 panel QS CAQ 11.1 [6.0, 19.7] 109 weighted
2018 2018 panel QS PCQ n.l. 109 weighted
2018 2018 panel QS Other 3.0 [0.7, 11.4] 109 weighted
2018 2018 panel QS Did not vote 8.5 [4.1, 16.9] 109 weighted
2018 2018 panel CAQ PLQ 2.5 [0.8, 7.4] 219 weighted
2018 2018 panel CAQ PQ 0.0 219 weighted
2018 2018 panel CAQ ADQ — 219 weighted
2018 2018 panel CAQ QS 0.5 [0.1, 3.2] 219 weighted
2018 2018 panel CAQ CAQ 85.5 [79.0, 90.3] 219 weighted
2018 2018 panel CAQ PCQ n.l. 219 weighted
2018 2018 panel CAQ Other 0.9 [0.3, 3.0] 219 weighted
2018 2018 panel CAQ Did not vote 10.6 [6.6, 16.5] 219 weighted
2018 2018 panel Other PLQ 0.0 34 weighted
2018 2018 panel Other PQ 0.0 34 weighted
2018 2018 panel Other ADQ — 34 weighted
2018 2018 panel Other QS 3.7 [0.9, 14.1] 34 weighted
2018 2018 panel Other CAQ 0.0 34 weighted
2018 2018 panel Other PCQ n.l. 34 weighted
2018 2018 panel Other Other 85.3 [69.5, 93.7] 34 weighted
2018 2018 panel Other Did not vote 11.0 [4.0, 26.6] 34 weighted
2018 2018 panel None / would not vote PLQ 1.6 [0.2, 10.6] 37 weighted
2018 2018 panel None / would not vote PQ 2.0 [0.3, 13.2] 37 weighted
2018 2018 panel None / would not vote ADQ — 37 weighted
2018 2018 panel None / would not vote QS 4.2 [1.0, 15.8] 37 weighted
2018 2018 panel None / would not vote CAQ 22.0 [10.0, 41.8] 37 weighted
2018 2018 panel None / would not vote PCQ n.l. 37 weighted
2018 2018 panel None / would not vote Other 2.8 [0.6, 11.2] 37 weighted
2018 2018 panel None / would not vote Did not vote 67.4 [48.4, 82.0] 37 weighted
2018 2018 panel Undecided PLQ 13.4 [7.0, 24.1] 75 weighted
2018 2018 panel Undecided PQ 7.9 [3.5, 17.1] 75 weighted
2018 2018 panel Undecided ADQ — 75 weighted
2018 2018 panel Undecided QS 9.0 [3.2, 22.8] 75 weighted
2018 2018 panel Undecided CAQ 36.9 [24.9, 50.8] 75 weighted
2018 2018 panel Undecided PCQ n.l. 75 weighted
2018 2018 panel Undecided Other 5.9 [2.3, 14.5] 75 weighted
2018 2018 panel Undecided Did not vote 26.7 [16.4, 40.4] 75 weighted
2022 QES 2022 PLQ PLQ 78.3 [67.4, 86.3] 126 weighted
2022 QES 2022 PLQ PQ 3.7 [1.4, 9.6] 126 weighted
2022 QES 2022 PLQ ADQ — 126 weighted
2022 QES 2022 PLQ QS 6.8 [2.9, 15.3] 126 weighted
2022 QES 2022 PLQ CAQ 1.6 [0.6, 4.1] 126 weighted
2022 QES 2022 PLQ PCQ 1.1 [0.3, 4.6] 126 weighted
2022 QES 2022 PLQ Other 0.4 [0.1, 2.8] 126 weighted
2022 QES 2022 PLQ Did not vote 7.9 [3.6, 16.8] 126 weighted
2022 QES 2022 PQ PLQ 2.1 [0.7, 6.0] 152 weighted
2022 QES 2022 PQ PQ 81.5 [72.7, 88.0] 152 weighted
2022 QES 2022 PQ ADQ — 152 weighted
2022 QES 2022 PQ QS 5.2 [2.2, 12.0] 152 weighted
2022 QES 2022 PQ CAQ 4.0 [1.7, 9.4] 152 weighted
2022 QES 2022 PQ PCQ 1.2 [0.2, 7.9] 152 weighted
2022 QES 2022 PQ Other 0.0 152 weighted
2022 QES 2022 PQ Did not vote 5.9 [2.4, 13.7] 152 weighted
2022 QES 2022 QS PLQ 11.2 [4.8, 23.9] 221 weighted
2022 QES 2022 QS PQ 8.4 [5.3, 13.1] 221 weighted
2022 QES 2022 QS ADQ — 221 weighted
2022 QES 2022 QS QS 66.8 [55.9, 76.2] 221 weighted
2022 QES 2022 QS CAQ 2.2 [0.7, 6.6] 221 weighted
2022 QES 2022 QS PCQ 0.9 [0.2, 3.7] 221 weighted
2022 QES 2022 QS Other 4.2 [0.7, 21.6] 221 weighted
2022 QES 2022 QS Did not vote 6.3 [3.4, 11.3] 221 weighted
2022 QES 2022 CAQ PLQ 1.9 [0.8, 4.6] 379 weighted
2022 QES 2022 CAQ PQ 6.0 [4.0, 9.0] 379 weighted
2022 QES 2022 CAQ ADQ — 379 weighted
2022 QES 2022 CAQ QS 2.6 [1.4, 5.0] 379 weighted
2022 QES 2022 CAQ CAQ 83.0 [78.0, 87.0] 379 weighted
2022 QES 2022 CAQ PCQ 0.3 [0.0, 2.0] 379 weighted
2022 QES 2022 CAQ Other 0.1 [0.0, 1.0] 379 weighted
2022 QES 2022 CAQ Did not vote 6.1 [3.7, 10.1] 379 weighted
2022 QES 2022 PCQ PLQ 6.2 [2.7, 13.4] 172 weighted
2022 QES 2022 PCQ PQ 2.0 [0.7, 5.5] 172 weighted
2022 QES 2022 PCQ ADQ — 172 weighted
2022 QES 2022 PCQ QS 2.3 [0.7, 7.9] 172 weighted
2022 QES 2022 PCQ CAQ 4.0 [1.9, 8.5] 172 weighted
2022 QES 2022 PCQ PCQ 71.8 [60.7, 80.8] 172 weighted
2022 QES 2022 PCQ Other 0.8 [0.2, 3.5] 172 weighted
2022 QES 2022 PCQ Did not vote 12.8 [5.8, 25.9] 172 weighted
2022 QES 2022 Other PLQ n < 30 27 weighted
2022 QES 2022 Other PQ n < 30 27 weighted
2022 QES 2022 Other ADQ — 27 weighted
2022 QES 2022 Other QS n < 30 27 weighted
2022 QES 2022 Other CAQ n < 30 27 weighted
2022 QES 2022 Other PCQ n < 30 27 weighted
2022 QES 2022 Other Other n < 30 27 weighted
2022 QES 2022 Other Did not vote n < 30 27 weighted
2022 QES 2022 Undecided PLQ 16.2 [9.0, 27.4] 79 weighted
2022 QES 2022 Undecided PQ 9.8 [5.0, 18.4] 79 weighted
2022 QES 2022 Undecided ADQ — 79 weighted
2022 QES 2022 Undecided QS 13.4 [7.2, 23.5] 79 weighted
2022 QES 2022 Undecided CAQ 13.1 [7.2, 22.5] 79 weighted
2022 QES 2022 Undecided PCQ 11.7 [5.3, 24.0] 79 weighted
2022 QES 2022 Undecided Other 12.3 [5.4, 25.5] 79 weighted
2022 QES 2022 Undecided Did not vote 23.6 [14.2, 36.6] 79 weighted

The diagonal dominates. In 2022, 83% of those who intended to vote CAQ reported a CAQ vote (a share that counts those who did not vote), and the flows between parties were small. Québec solidaire (QS) kept the smallest share of its campaign supporters, 67%: 11% of them voted PLQ and 8% PQ. In 2018, the picture is quite different. The CAQ drew from every side, with 12% of those who intended to vote PQ, 11% of those who intended to vote QS and 22% of those who had said they would not vote. In other words, the winner of 2018 did not only hold its voters; it kept recruiting during the campaign.

Taken together, the share of voters who did not vote as they had said is far from trivial: it ranges from 22% to 28% across the five panels. The largest group among them switched from one party to another, from 10% to 21% of voters, while the undecided and those who had said they would not vote made up the rest. In 2018 the two groups were about the same size (10% switchers, 10% undecided). That panel, however, did not ask the follow-up question of some undecided respondents interviewed by telephone, nor of those who said they would not vote, so it counts more undecided than the others. The 2007 election, a three-way race between the PLQ, the PQ and the Action démocratique du Québec (ADQ), produced the most switchers, 21%.

Where the undecided went

17% to 27% of the campaign's undecided did not vote; in 2018, 37% of them voted CAQ, the winnerWhat the respondents still undecided during the campaign reported after the election, by panel study
100% stacked horizontal bars, one per panel study from 1998 to 2022: what the respondents who were still undecided during the campaign reported after the election, with those who did not vote first. Between 17% and 27% did not vote; in 2018, 37% voted CAQ; in 2022, 24% did not vote and 13% voted CAQ. Values in the table view.100% stacked horizontal bars, one per panel study from 1998 to 2022: what the respondents who were still undecided during the campaign reported after the election, with those who did not vote first. Between 17% and 27% did not vote; in 2018, 37% voted CAQ; in 2022, 24% did not vote and 13% voted CAQ. Values in the table view.

Source: qesR. The undecided are respondents who named no party during the campaign, even when asked (where the panel asked) which party they leaned to, and who answered after the election. Weighted (2018 panel, QES 2022) or unweighted, no validated weight (1998, 2007 and 2012 panels). Outlined segment: did not vote. Numbers: shares of 8% or more. Each row rests on 104 respondents or fewer, so the confidence intervals in the table view are wide. In the 2007 and 2018 panels, 'don't know' and 'refused' cannot be told apart and are both counted as undecided.

Table view
Election Study Reported vote Share of the undecided, % [95% CI] n Weighting
1998 1998 polls PLQ 38.3 [28.4, 49.3] 81 unweighted: no validated weight
1998 1998 polls PQ 28.4 [19.7, 39.1] 81 unweighted: no validated weight
1998 1998 polls ADQ 13.6 [7.7, 22.9] 81 unweighted: no validated weight
1998 1998 polls QS — 81 unweighted: no validated weight
1998 1998 polls CAQ — 81 unweighted: no validated weight
1998 1998 polls PCQ — 81 unweighted: no validated weight
1998 1998 polls Other 2.5 [0.6, 9.3] 81 unweighted: no validated weight
1998 1998 polls Did not vote 17.3 [10.5, 27.1] 81 unweighted: no validated weight
2007 2007 panel PLQ 19.2 [12.8, 27.9] 104 unweighted: no validated weight
2007 2007 panel PQ 26.9 [19.3, 36.2] 104 unweighted: no validated weight
2007 2007 panel ADQ 23.1 [16.0, 32.1] 104 unweighted: no validated weight
2007 2007 panel QS 0.0 104 unweighted: no validated weight
2007 2007 panel CAQ — 104 unweighted: no validated weight
2007 2007 panel PCQ — 104 unweighted: no validated weight
2007 2007 panel Other 4.8 [2.0, 11.0] 104 unweighted: no validated weight
2007 2007 panel Did not vote 26.0 [18.4, 35.2] 104 unweighted: no validated weight
2012 2012 panel PLQ 29.5 [19.4, 42.1] 61 unweighted: no validated weight
2012 2012 panel PQ 23.0 [14.1, 35.1] 61 unweighted: no validated weight
2012 2012 panel ADQ — 61 unweighted: no validated weight
2012 2012 panel QS 3.3 [0.8, 12.2] 61 unweighted: no validated weight
2012 2012 panel CAQ 14.8 [7.9, 26.0] 61 unweighted: no validated weight
2012 2012 panel PCQ n.l. 61 unweighted: no validated weight
2012 2012 panel Other 4.9 [1.6, 14.2] 61 unweighted: no validated weight
2012 2012 panel Did not vote 24.6 [15.4, 36.9] 61 unweighted: no validated weight
2018 2018 panel PLQ 13.4 [7.0, 24.1] 75 weighted
2018 2018 panel PQ 7.9 [3.5, 17.1] 75 weighted
2018 2018 panel ADQ — 75 weighted
2018 2018 panel QS 9.0 [3.2, 22.8] 75 weighted
2018 2018 panel CAQ 36.9 [24.9, 50.8] 75 weighted
2018 2018 panel PCQ n.l. 75 weighted
2018 2018 panel Other 5.9 [2.3, 14.5] 75 weighted
2018 2018 panel Did not vote 26.7 [16.4, 40.4] 75 weighted
2022 QES 2022 PLQ 16.2 [9.0, 27.4] 79 weighted
2022 QES 2022 PQ 9.8 [5.0, 18.4] 79 weighted
2022 QES 2022 ADQ — 79 weighted
2022 QES 2022 QS 13.4 [7.2, 23.5] 79 weighted
2022 QES 2022 CAQ 13.1 [7.2, 22.5] 79 weighted
2022 QES 2022 PCQ 11.7 [5.3, 24.0] 79 weighted
2022 QES 2022 Other 12.3 [5.4, 25.5] 79 weighted
2022 QES 2022 Did not vote 23.6 [14.2, 36.6] 79 weighted

The undecided matter less for the result than for turnout. Between 17% and 27% of them did not vote, compared to 7% to 14% of those who had named a party. Voters who decide during the campaign are known to be the most open to its effects (Fournier et al. 2004); in Quebec, a sizeable minority of the undecided simply stay home. Those who did vote went in no single direction. In 1998 they leaned to the PLQ (38% versus 28% for the PQ), but on only 81 respondents. In 2018 they broke for the CAQ, and in 2022 they split across all the parties. The idea that the undecided lean toward one party finds no steady support here.

Who won the campaign

The matrix shows who moved; it does not show who came out ahead once the moves in both directions are added up. We take, for each party, its share of the reported vote minus its share of the intentions, among the same respondents. A positive gap means the party won more voters during the campaign than it lost.

On net, the campaign moved a few points at most: from −3.1 pts for the PCQ in 2022 to +6.1 pts for the CAQ in 2018Each party's share of the reported vote minus its share of the campaign intentions, among the same respondents, with 95% confidence intervals
Dot chart in four panels, one per panel study from 2007 to 2022, one row per party: the party's share of the reported vote minus its share of the campaign intentions, among the same respondents, in points, with confidence intervals. Most gaps are within a few points of zero. The largest gain is that of the CAQ in 2018, +6.1 pts, the largest loss that of the PCQ in 2022, −3.1 pts; in 2022 the CAQ is at −2.9 pts and the PQ at +3.0 pts. Values in the table view.Dot chart in four panels, one per panel study from 2007 to 2022, one row per party: the party's share of the reported vote minus its share of the campaign intentions, among the same respondents, in points, with confidence intervals. Most gaps are within a few points of zero. The largest gain is that of the CAQ in 2018, +6.1 pts, the largest loss that of the PCQ in 2022, −3.1 pts; in 2022 the CAQ is at −2.9 pts and the PQ at +3.0 pts. Values in the table view.

Source: qesR. All respondents interviewed during the campaign and after the election, including the undecided, those who said they would not vote and those who did not vote, so that the gains of the parties and of abstention add up. Filled dots: weighted with the post-election weight (2018 panel, QES 2022); hollow dots: unweighted, no validated weight (2007 and 2012 panels). The 1998 polls are left out: their recontact over-represents the undecided, non-disclosers and third-party (ADQ) supporters, so their unweighted gains cannot be read (see the text). Parties that did not run, or that the question did not list, are left out.

Table view
Election Study Party Net gain, points [95% CI] n Weighting
2007 2007 panel PLQ -2.4 [-4.2, -0.7] 1477 unweighted: no validated weight
2007 2007 panel PQ -0.9 [-3.0, 1.2] 1477 unweighted: no validated weight
2007 2007 panel ADQ 1.4 [-0.6, 3.4] 1477 unweighted: no validated weight
2007 2007 panel QS -2.1 [-3.1, -1.1] 1477 unweighted: no validated weight
2012 2012 panel PLQ 1.0 [-1.2, 3.3] 689 unweighted: no validated weight
2012 2012 panel PQ 4.6 [2.2, 7.1] 689 unweighted: no validated weight
2012 2012 panel QS -0.4 [-2.2, 1.3] 689 unweighted: no validated weight
2012 2012 panel CAQ -2.6 [-5.1, -0.2] 689 unweighted: no validated weight
2018 2018 panel PLQ -2.9 [-5.0, -0.8] 815 weighted
2018 2018 panel PQ -2.2 [-4.8, 0.3] 815 weighted
2018 2018 panel QS -2.5 [-4.5, -0.5] 815 weighted
2018 2018 panel CAQ 6.1 [3.0, 9.3] 815 weighted
2022 QES 2022 PLQ 2.3 [-0.3, 4.8] 1156 weighted
2022 QES 2022 PQ 3.0 [1.4, 4.7] 1156 weighted
2022 QES 2022 QS -2.2 [-5.0, 0.7] 1156 weighted
2022 QES 2022 CAQ -2.9 [-4.7, -1.1] 1156 weighted
2022 QES 2022 PCQ -3.1 [-5.2, -1.0] 1156 weighted

Most gaps are small, within a few points of zero. This is what one would expect when most voters keep their intention and the remaining moves partly cancel out. The two largest gains went to parties that went on to form the government: the CAQ in 2018, with 6.1 points (95% confidence interval: 3.0 to 9.3), and the PQ in 2012, with 4.6 points (2.2 to 7.1). The 2018 gain fits Durand and Blais’s (2020) account. Winning the government did not require winning the campaign. The PLQ in 2007 and the CAQ in 2022 formed the government while losing ground during the campaign (by 2.4 and 2.9 points), and the PQ gained 3.0 points in 2022 without forming it.

We leave the 1998 polls out of the figure, because their net gains cannot be read as they stand. The 1998 panel in qesR is made of two of the three polls that Durand and Blais (1999) recontacted, those of CROP and CREATEC, which interviewed francophones only. The recontact was drawn to over-represent the undecided, non-disclosers and supporters of third parties, and it is analysed here without weights (they are not validated; see the weights table). Unweighted, those who intended to vote ADQ make up 23% of the pairs; with the weight the producers used for their own published tables, they make up 16%. The unweighted numbers show the PLQ and the PQ gaining about as much (2.7 and 2.3 points), so there is no Liberal swing at the PQ’s expense, and the ADQ losing 6.2 points. With the producers’ weight, the PLQ’s gain is close to zero (0.3 points) and the PQ’s loss is small (1.5 points), both with confidence intervals that include zero, while the ADQ’s loss shrinks to 3.2 points. Either way, these numbers are consistent with Durand and Blais (1999). This is hardly surprising: they come from the same respondents, so they confirm the original analysis rather than test it independently.

What the numbers mean and their limits

Overall, we find mixed support for the view that Quebec elections are decided at the last minute. On the one hand, from 22% to 28% of voters did not vote as they had said, and the CAQ of 2018 gained ground during the campaign itself. On the other hand, most of these moves cancel out, the undecided are few and abstain more often than the other voters, and the net effect of the campaign on a party’s share is usually a few points. Two limits of these data call for caution.

First, the panels see only those who stayed. Respondents who dropped out after the campaign are not in these tables, and if leaving the panel is related to changing one’s mind, the transitions lean toward stability. Second, the vote after the election is a vote that respondents report once the winner is known. Some voters may recall a vote closer to the result than the one they cast, which would inflate the gains of the winner, the CAQ in 2018 included. If recall simply leaned toward the winner, though, the party that formed the government would gain every time, and it lost ground in 2007 and 2022. The five studies also differ in design and mode, and three of them are analysed without weights, so comparisons across elections are best read as orders of magnitude.

That most voters keep their intention across five elections and several survey designs is itself a finding. Of course, this stability may not survive more fragmented elections, with five parties above 10% of the vote as in 2022; that is a question the next panels will have to answer.

Going further

  • Durand, Claire, and André Blais. 1999. “Why Did the Polls Go Wrong in the 1998 Quebec Election? The Answer from Post-Election Polls.” Bulletin of Sociological Methodology / Bulletin de Méthodologie Sociologique 62 (1): 43–47. https://doi.org/10.1177/075910639906200105
  • Durand, Claire, and André Blais. 2020. “Quebec 2018: A Failure of the Polls?” Canadian Journal of Political Science 53 (1): 133–150. https://doi.org/10.1017/S0008423919000787
  • Fournier, Patrick, Richard Nadeau, André Blais, Elisabeth Gidengil, and Neil Nevitte. 2004. “Time-of-Voting Decision and Susceptibility to Campaign Effects.” Electoral Studies 23 (4): 661–681. https://doi.org/10.1016/j.electstud.2003.09.001
  • Johnston, Richard, André Blais, Henry E. Brady, and Jean Crête. 1992. Letting the People Decide: Dynamics of a Canadian Election. Montreal and Kingston: McGill-Queen’s University Press. https://doi.org/10.1515/9780773563643
  • On the PCQ voters of 2022, from the same Quebec Election Study: Bélanger, Éric, Philippe Mongrain, Thomas Gareau-Paquette, and Valérie-Anne Mahéo. 2025. “A Party that Went Viral? The Drivers of Support for the Parti Conservateur du Québec in the 2022 Election.” Canadian Journal of Political Science 58 (2): 277–296. https://doi.org/10.1017/S0008423924000829

About the data

  • Studies. The studies that interviewed the same respondents during the campaign and after the election: the 1998 CROP and CREATEC polls (francophones only), the panels Claire Durand ran in 2007, 2012 and 2018, and the 2022 Quebec Election Study. Only respondents of both waves are used. The 2022 study was an online panel, the 2018 panel mixed telephone and web, and the older studies were by telephone.
  • Intention. The party named during the campaign. Respondents who hesitated were asked which party they leaned to, and that answer counts; those who still named no party are the undecided. In the 2018 panel, some undecided respondents interviewed by telephone and those who said they would not vote were not asked the follow-up, so that panel counts more undecided. In the 2007 panel, a few respondents keep their first answer, where the producer’s combined answer cannot be used.
  • Vote. The party reported after the election, or did not vote.
  • Weights. The post-election weight of each respondent in the 2018 panel and the 2022 study. The 1998, 2007 and 2012 panels have no validated weight (see the weights table): they are unweighted here and drawn hollow. The 1998 recontact also over-represents the undecided, non-disclosers and third-party (ADQ) supporters, so its net gains are left out of the last figure; the text compares them with and without the weight the producers used.
  • Attrition. Respondents who left the panel after the campaign are not in these tables.
  • Code. Every figure starts from the pooled vote_choice, in the long layout (one row per respondent and wave), matched within each study.
The call that builds the data 
h <- qes_harmonize(
  studies = c("qes1998", "qes2007_panel", "qes2012_panel", "qes2018_panel", "qes2022"),
  targets = "vote_choice",
  layout = "long", missing = "reasons", quiet = TRUE
)