The pollsters pretty much got it right for the May 2017 BC provincial election. Look at the final polls from the bigger polling firms:
Forum Research BCL 39 NDP 41 Green 17
Insights West BCL 41 NDP 41 Green 17
Mainstreet BCL 39 NDP 40 Green 20
Ipsos Reid BCL 39 NDP 40 Green 17
Angus Reid BCL 40 NDP 41 Green 15
Actual Results BCL 41 NDP 41 Green 17
With the exception of Mainstreet pegging Green support 3% higher than the actual results, all the other numbers were within 0-2% of the final results and well within the margins of error. A huge success for pollsters, especially since they (in the opinion of many voters and media pundits) “got it wrong” in predicting the 2013 BC election.
But did they get it wrong? It turns out that they may not have.
Polls of voting intention aren’t good at predicting an exact result because polls are a snapshot in time, and out of necessity are done in advance of election day. Where polls are most useful is in trend analysis. If the polls barely budge throughout the campaign, you might expect that to predict pretty well where the parties will end up. When the polls show consistent momentum for one party throughout the campaign, you might expect more people to switch their vote at the last minute, and that the party with the momentum may get more than their fair share of the undecided voters. Similarly, if a party consistently goes down during the campaign you might expect them to continue to do so on election day. In this regard, the polls did help in 2013. Below is the percentage reported for the BC Liberals from all the pollsters that publicly released two or more polls during the 2013 election:
EKOS 28.4% 34.5%
Angus Reid 28% 31% 34% 36% 36%
Ipsos Reid 29% 35% 37% 37%
Justason 27% 31%
Forum 35% 41%
These polls were all taken at different times of the campaign, which partly explains the different numbers. However, the important thing to observe is that Liberal support stayed the same or went up in every single new poll. The Liberals never once dropped in any poll throughout the entire campaign. The pollsters disagreed about their exact levels of support, but agreed about the trend. And that trend continued after the polling stopped, with more people continuing to move to the BC Liberals in the final days of the 2013 campaign, and on election day itself. So again, the polls weren’t wrong, the trend the polls identified simply continued in the gap between the final polls and election day.
So I understand scepticism about polls predicting exact results, but I don’t look to them for that. I look to them for trend analysis, and from that perspective they are surprisingly reliable.