Monday, April 29, 2013

Forum Ontario Poll: 36% OLP, 36% PC, 24% NDP

Following recent polls from EKOS and Ipsos Reid, Forum Research has released its monthly update for Ontario's provincial scene. Alongside the release came a hilarious overreaction on the part of certain person whose book I struggled to finish and now wish I had my money and my time back from, but the less said about them the better. Just go read this.

Forum (Ontario Provincial - April 26th, 2013 - +/- 3%)
Ontario Liberal Party: 36% (+3%) - 52 seats (+14 seats)
Prog. Conservative: 36% (+1%) - 42 seats (-6 seats)
New Democratic: 24% (-2%) - 13 seats (-8 seats)
Green Party: 4% (-1%)

Compared to the previous Forum poll out in March, the Liberals have jumped up seemingly at the expense of the New Democrats. However its clear the Liberals are not out of the woods yet, as they continue to be tied up with the PCs for the lead.

Given the recent polls from EKOS and Ipsos that seem so different, can we say that the current polls right now are volatile? Who can we say is right or wrong? We'll discuss that in a bit. Lets look at the regional numbers first.

The Liberal's lead in seats come from their domination in the GTA, including 47% in the 416 (Toronto) and a closely fought race in the 905 Region (34% to 37% for the PCs).  However, if you're looking for a reason why the Liberal's numbers are bolstered, look towards Southern Ontario. Something we haven't seen often since 2011 is a very close race between the PCs and Liberals (36% to 33%), with the NDP sititng far back. These are numbers broadly similar to the 2011 election results in the region, (42% PC, 32% Lib, 21.5% NDP), meaning the Liberals actually manage to hold their Southwestern Ontario seats, and even gain a couple. Those seats are very crucial if the Liberals want to hold government, as sadly you're not going to build a majority or major-minority out of the GTA alone.

The Liberals also have a lead in Northern Ontario (37% to 33% PC and 25% NDP, but the North has a small sample size so be wary), while the PCs lead in Eastern Ontario as well (48% to 28% Liberal). I'll have a map up at the end of this post.

Whats Up With These Polls?

As I said before, EKOS, Ipsos, and Forum are all giving us different numbers of a sort. Why is this? Who is right? Is there even a trend? All questions that I'll try to answer quickly, but first here are the last four polls done this month (including the end of March):

Forum (Mar 26-27): 35% - 33% - 26% - 5%
EKOS (April 3-10): 32% - 31% - 26% - 10% (rounded)
EKOS (Likely Voters): 38% - 27% - 27% - 6% (rounded)
Ipsos (April 12-17): 37% - 29% - 28% - n/a
Forum (April 26): 36% - 36% - 24% - 4%

If you're perceptive, you'll notice two things right away. One is that while EKOS and Forum share a broadly similar trend (close PC-Lib races) in their topline numbers, when EKOS applied its "Likely Voter" screen, their numbers instead looked more similar to Ipsos Reid's. The second is that Forum stayed consistent within its own trend - there was no wild swing between their March release and today's.

While we can't say for certain, its clear that a direct comparison between the Ipsos/EKOS' Likely Voters, and Forum/EKOS' topline may be somewhat like comparing apples and oranges. I believe Ipsos has a methodology which incorporates their Likely Voter model into their topline; EKOS' was nice enough to provide both versions, while I can't say what Forums' methodology nitty-gritty is like. If Forum had shown Likely Voters, I suspect the numbers we see now would've been much different. So, there exists that major fault line that makes comparison difficult. The second is that both EKOS and Forum used IVR, while Ipsos used an online survey.

So, that is my basic guess as to why we're getting different numbers. If we had some other pollsters enter the field, such as Angus Reid or Leger, we could probably get an even clearer picture.

The question then becomes who is the most trustworthy? Despite the inane hatred some feel for Forum, their final poll for Ontario's 2011 election was more accurate then either Ipsos or EKOS - in fact it was almost spot on. However, Forum has had flops like Alberta and Quebec - but so did EKOS, and Ipsos isn't 100% accurate either. The records of all polling companies is hit or miss, that is just how the industry goes. Take every poll with a grain of salt, because they will get it wrong eventually, there is no avoiding a bad call at some point.

But, in the context of the present (and in the lead up to the next provincial election), who should we put more stock in? Who will have the more accurate trend?

Well, in the run up to E-day, legitimate pollsters will tend to come together, so don't worry about it then as much. Pre-writ polls like the ones we're discussing here are trickier, because they're gauging opinions of respondents outside of the context of an election. People are more undecided and fluid about their choices, because hey, we could be two years away from an election, most aren't even thinking about these issues (hence why pollsters tend to go crazy in the aftermath of big political stories, their samples are more likely to be paying attention; monthly pollsters like Nanos or Forum tend to poll during the interim period a lot of times).

So really, we can't say who is going to be more accurate on their own. All we can say is that there is a general trend, and that trend generally tends to be accurate in a broad sense. What it comes down to is whether you believe the Likely Voter models given by EKOS and (probably) Ipsos are more accurate then your basic topline numbers. Personally I'm not convinced their Likely Voter models are more accurate than what Forum or EKOS' topline shows. Hence why I don't put into my model the Likely Voter numbers EKOS gave - I'm not sure whether or not its going to be accurate (not even EKOS is sure), so I'll stick with my traditional pollsters. Ipsos gets away with it because they don't have any alternate numbers.

But that is my personal choice, and others may believe the Likely Voter models are more accurate. The key thing is not to presuppose accuracy or inaccuracy on the pollster's part, like some do. Presupposing that a pollster is inaccurate is not only going to lead you to wrong conclusions, its the hallmark of an intellectually dishonest person, in my opinion. What you need to look at is the broader trend, the methodology, and so on. You need evidence to make a claim, you can't just assume its true.

Speaking of trends, is there one to be found? Well, we only have four polls so our sample is kind of small, but even with three we can make a broad trend out of it. Its clear that the PCs are either in third or tied within the margin of error with the Liberals, except for Ipsos which shows them with a large lead, and the NDP sit between the mid-to-high twenties. The big caveat is the Liberal numbers, which range from mid thirties to high twenties, a fairly big gap - but as we noted, this is seemingly more due to the presence of Likely Voter models, which is a big caveat when discussing any trend out of these polls.

Phew, anyways, that is my bit on this. Hopefully I've cleared up some misconceptions and questions. Not that anyone asked, I just figured they're out there. Here's a map:

1 comment: