I’ve written before about how polls ought to carry warning labels urging people to read with caution. The two polls we now have for the upcoming territorial election are no different.
But even though we should take care drawing conclusions from these results doesn’t mean they are meaningless. It just means thinking critically about the numbers and consider them within the context of known limitations about polling methodology.
We’ll start with the poll released late Friday purporting to show the Liberals at 46 per cent, with the Yukon Party and NDP trailing at 30 per cent and 24 per cent respectively. The poll was quickly dismissed by some because it was commissioned by the Liberals and conducted by a company with deep ties to Liberals elsewhere in the country.
These criticisms are not necessarily well founded. It is possible that the Liberals conducted several polls and only chose to release the one that was most favourable to them in an attempt to siphon away anti-Yukon Party votes from the NDP. But bear in mind that the Liberals aren’t terribly well-funded — according to most recent fundraising numbers at least — and conducting multiple polls is by no means cheap.
The way the questions were asked doesn’t seem to have been designed to lead people to pick one option over the others.
So while the decision on the part of the Liberals to release the poll was undeniably political, there isn’t much reason to think that the partisan nature of the poll is cause to reject it as one piece of evidence about the state of the race.
The poll lumped in undecided voters which is not an uncommon practice. But it is pretty easy for anyone to do some basic calculations to pull the undecideds out of the totals. By my math, about 40 per cent of respondents favour the Liberals, 26 per cent back the Yukon Party and the NDP has 21 per cent, with 13 per cent of voters undecided.
So what should we take from the poll? Before we draw any conclusions we have to consider the margin of error.
Rather than thinking that polls show a precise level of partisan support, it’s more helpful to think of their findings as a range of possibilities about what the true number is. The Gandalf poll came with a 4.9 per cent margin of error. This means that (if we separate out the undecideds) the Liberals are somewhere between 35 and 45 per cent, the Yukon Party between 21 and 31 per cent, the NDP between 16 and 26 per cent, and somewhere between 8 and 18 per cent of the electorate is undecided.
Note that there is no overlap between Liberals and the other parties. This means that, subject to a couple of caveats, this poll would seem to a real Liberal lead, at least at the time the it was conducted.
The caveats? First, like all polls, this one claims to be accurate 19 times out of 20. That means that about 5 per cent of the time the actual numbers could be outside of that range.
Second, all of this assumes that Gandalf drew a random sample from the population, and didn’t disproportionately draw people of a particular political persuasion.
A common polling flaw is when pollsters only call landlines — disproportionally maintained by older voters, while many younger voters only have cell phones. But Gandalf did its poll by calling half landlines and half cell phones, which isn’t too far from CRTC numbers about types of phone ownership.
Meanwhile, a second poll was released by Yukon based Datapath Systems.
The first thing that jumps out at me about this poll is that Datapath didn’t actually ask the question that matters the most on Election Day: who respondents intend to vote for.
Instead Datapath asked people who they would like to see as premier with 37 per cent of respondents picking Sandy Silver compared to 34 per cent for Darrell Pasloski and 27 per cent for Liz Hanson. They also asked which candidate people would like to see win in their riding. On that question, the Liberals and Yukon Party each got 34 per cent and 29 per cent picked the NDP.
Since people base their vote on a number of considerations, including the leader, party brand, platform and local candidate, it is hard to draw a clear line from the results of the Datapath poll to a particular vote. As someone who tends to vote based on leader and party, I’m inclined towards the first question, but there are many people for whom the local candidate is essential.
There is often a lot of criticism of Datapath’s polling for the fact that their samples are not truly random and there is a certain amount of self-selection by respondents through their online approach. But in their defense, Datapath does weigh numbers based on the age, gender, and race of the population at large, which helps correct for some of those flaws.
Contrary to popular belief, Datapath actually has a pretty decent track record of predicting results within the margin of error. The problem was that people who read previous Datapath polls in the past committed the first error of poll reading and assumed that the numbers they produced were exact figures, rather than a range and were surprised when the ultimate numbers were three, four, even five points away.
It is difficult to do a direct comparison between the two polls since Datapath’s poll didn’t actually ask people who they intended to vote for. But it is worth noting that they are consistent with one another in having the Liberals in or near the lead and the NDP trailing.
There is still significant uncertainty in this race however and things can change a lot before election day. The Gandalf poll showed that 13 per cent of voters hadn’t made up their mind yet. Which way those voters break could tip the election for any of the parties.
Among the rest of the voters, many of the respondents were simply “leaning” which means that they haven’t fully made up their mind yet. According to Gandalf, only 62 per cent of the electorate is fully decided.
And even if the numbers are correct, we don’t know how it will all work out on a riding-by-riding basis. After all, you can win the popular vote but lose the election if your votes are too concentrated in certain ridings.
We’ll know soon enough which of these polling approaches came closest to a prediction. Or if they’re both wildly off the mark.
Kyle Carruthers is a born-and-raised Yukoner who lives and practises law in Whitehorse.