Four reasons why the Polls could be wrong in Canada
Major takeaways from a combination of Google Search and polling data.
- Liberals will lose some share
- NDP will gain share but not much
- Conservatives will perform close to the 2019 share
- PPC is hurting all parties and not just the Tores
- Bloc will underperform 2019
- Reasons why the Polls could be wrong in Canada: They are probably overstating the NDP
- Reasons why the Polls could be wrong in Canada: Polls have huge variance in province level data
- Reasons why the Polls could be wrong in Canada: Polls are generally wrong
- Many Canadians are yet to make up their mind on who to vote
Reasons why the Polls could be wrong in Canada: They are probably overstating the NDP
- Daily tracking polls show NDP at 18.1% % while Weekly tracking polls show NDP at 20.7%. Assuming one of the two is wrong, the results will look quite different given how competitive the election has been so far.
- But there is an additional problem. Google Search data shows NDP has underperformed versus the 2019 election. This does not tie in with data reported by all the polling firms. What we know for sure is that polling firms over reported NDP share in 2019. Our hypothesis is that the younger voter base of NDP are more open to answering polls
Reasons why the Polls could be wrong in Canada: Polls have huge variance in province level data
Let us take the heavily contested province of Quebec. The gap between LPC and CPC is 15% according to Leger but only 8% according to Ekos. While the gap between LPC and the Bloc is 4% according to Leger and Ekos, Mainstreet claims the gap is just 0.3%. Abacus has a 19% gap between LPC and CPC and a 7% gap between LPC and the Bloc. Given the competitiveness of the election, variances of 3-4% can produce very different results when t comes to seat forecasts.
Reasons why the Polls could be wrong in Canada: Polls are generally wrong
We track every poll on the planet and one can say with utmost confidence that pre-election polls are good at picking the direction but not so great in picking the exact vote or seat shares.
In the recently concluded California recall and Boston Mayoral Prelim election, Google search data was more accurate in predicting a landslide for Gavin Newsom or that Annissa George would finish second in the Boston race
For example, in the 2019 election, the respected 338, missed the Liberal seat forecasts by 10%. However, 338 over-estimated the New Democrats by 46%. These are excellent estimates to be honest but the 10% under-estimation for Liberals meant that the gap between the Liberals was 36 seats and not 17 seats as estimated by 338.This happens in country after country – United States, Israel and New Zealand . Exit Polls are far more accurate and consistently across countries.
Many Canadians are yet to make up their mind on who to vote
Read our Coverage of the Canadian Election