This article aims to answer the most frequently asked questions about polls during the election campaign after providing a brief reminder of the last Quebec campaign.
Posted at 2:00 p.m.
What happened in 2018?
It will be recalled that the 2018 election campaign ended with poor predictions in the polls for the two leading parties, a situation that had not been seen in Quebec since the 1998 elections. The polls had estimated that the Parti liberal du Québec (PLQ) and the Coalition avenir Québec (CAQ) were neck and neck with between 30% and 35% of the voting intentions and that the Parti québécois (PQ) and Québec solidaire (QS) were neck and neck- Elbow, between 15% and 20%. However, the CAQ finished 13 points ahead of the PLQ. What happened ?
It was easy to conclude that the polls were “broken”, especially since several bad predictions had also taken place in other countries.
The situation was all the more astonishing since all the polls, using different methodologies and sample sources, had predicted the results badly, and this only for the two leading parties. It is therefore difficult to attribute the prediction error to the methodology. My colleague André Blais and I showed1 that the poor prediction could be explained by a last-minute move towards the CAQ on the part of voters from other parties.
Could we have foreseen this outcome? This is a question that researchers have been asking for a long time when certain conditions are met: the presence of several parties of equal strength and shifts in voting intentions during the campaign, among others. However, bad predictions have always turned out to be surprises.
What about this time?
At the start of the campaign, there is only one party in the lead in terms of voting intentions, the other parties finding themselves with similar voting intentions. However, movements that would see certain parties stand out at the end of the campaign have already occurred during recent campaigns, both federal and provincial. This is a possibility that should not be overlooked and it is the polls that will be able to tell us if this situation arises.
How do you know which pollster to trust? The scientific answer to this question is first to look at all the polls.
Pollsters do not all use the same methodology. When these methodologies converge to give similar voting intentions, we can have good hope that we have the right time. Moreover, the more a sonar gives precise and transparent information on its methodology and the characteristics of its sample, the more it can be trusted. Do some pollsters have a bias? Personally, I don’t think so, and for a simple reason. Pollsters stake their credibility in election polls. They have no interest in manipulating the results. Moreover, there is no indication that such manipulation can help a party.
How to read the polls?
It will be important to look at which social groups are attracted to which parties. A voting intention of 15% scattered throughout Quebec will not have the same impact as a voting intention concentrated in certain regions or certain age groups. However, it should be noted that the voting intentions published for the subgroups are subject to enormous margins of error – credibility intervals for web polls. And so the published numbers will usually vary widely from poll to poll for these subgroups. It should also be noted that participation is very variable according to the groups. Older people and Francophones vote more than young people and non-Francophones.
What are the consequences ? Aggregators – like Qc125, for example – try to predict how many seats each party will win. To do this, they rely on polls, in addition to estimates by region and other “historical” parameters such as the vote in each constituency in previous elections, the socio-demographic composition of constituencies, etc. As the voting intentions for the subgroups are more uncertain, so are their predictions.
Also, with five-party races in some cases, the uncertainty is very high.
If the voting intentions for the CAQ remain as they are now, the prediction will be more reliable for this party. However, for the other parties, particularly those with a predominantly francophone electorate, the uncertainty is very great.
Track last minute changes
It is desirable that several pollsters using different methodologies be present in this campaign, which would make it possible to have better confidence in the estimates if they converge. It would also be desirable for the polls to be carried out as late as possible in the campaign to be able to track last-minute changes and thus avoid surprises.