Federal Election Polling Thread

[Looch City]

The Conservative ad is working. People realized Harper isn't perfect but he's the best we've got right now.

[EldrickOnIce]

Originally Posted by Iggy City

The Conservative ad is working. People realized Harper isn't perfect but he's the best we've got right now.

[Bownesian]

Originally Posted by PeteMoss

And EKOS has the CPC ahead. So three polls.. three different parties in front.

That means they are tied and all this handwringing is over the natural statistical variance due to the margin of error. The trend is a near-tie with a slight NDP lead and no movement yet.

Also, anyone who claims that a single poll is more likely correct because its result shows a change beyond the MOE may well be misunderstanding the math. The margin of error reported in these polls is based on the error falling within that range 19/20 times, which means that there should be an outlier 1 out of 20 times. Basically an outlier may indicate a trend or it may be an outlier, but given enough time, outliers tend to be swamped the weight of more data.

I predict we won't have any real movement from this deadlock until several points in the campaign:

- the debates, each of which could produce movement, especially the next one which could crystallize some peoples' thoughts about the various foreign affairs issues that have come up in the campaign so far

- the release of full platforms (we haven't really heard the environmental GHG tax/regulations/emissions trading parts of the Liberal or NDP platforms, which may well prove to be unpalatable for industry and/or consumers if the past platforms have anything to say)

- Thanksgiving, when families get together and talk politics and finalize their decisions a week before the end of the campaign

[Dan02]

So what's off with the Forum polling that they always seem to be oversampling NDP supporters?? This is going all the way back to the last election too. They claim the smallest margin of error yet seem to be the most consistently off.

[Iowa_Flames_Fan]

Originally Posted by Bownesian

That means they are tied and all this handwringing is over the natural statistical variance due to the margin of error. The trend is a near-tie with a slight NDP lead and no movement yet.

Also, anyone who claims that a single poll is more likely correct because its result shows a change beyond the MOE may well be misunderstanding the math. The margin of error reported in these polls is based on the error falling within that range 19/20 times, which means that there should be an outlier 1 out of 20 times. Basically an outlier may indicate a trend or it may be an outlier, but given enough time, outliers tend to be swamped the weight of more data.

I predict we won't have any real movement from this deadlock until several points in the campaign:

- the debates, each of which could produce movement, especially the next one which could crystallize some peoples' thoughts about the various foreign affairs issues that have come up in the campaign so far

- the release of full platforms (we haven't really heard the environmental GHG tax/regulations/emissions trading parts of the Liberal or NDP platforms, which may well prove to be unpalatable for industry and/or consumers if the past platforms have anything to say)

- Thanksgiving, when families get together and talk politics and finalize their decisions a week before the end of the campaign

To me, one advantage of polling aggregators is that they will smooth out the natural variances you're talking about, but I suspect your assessment of the state of the race is about right.

However, it's worth stressing that the "19 times out of 20" principle doesn't explain Forum, which has had an NDP lead well outside the margin of error for three consecutive samples.

If we remove that one pollster, what the remaining data suggests is exactly what you've indicated--a close race, with perhaps a very narrow NDP lead. This is an unprecedented campaign in that it started long before most voters were interested in following an election. My guess is that voters start tubing in over the next two weeks or so, and that the needle won't move much until that happens.

[Bownesian]

Yeah, I can't explain Forum without dipping into conspiracy territory. How they could be as chronically off as they have been and not change their methodology is beyond explanation.

[Parallex]

Three new polls...

Nanos: http://www.nanosresearch.com/library...0TrackingE.pdf

Innovative: http://www.innovativeresearch.ca/sit...rated_Deck.pdf

Abacus: http://abacusdata.ca/desire-for-chan...ection-canada/

The Skinny: All three pollsters agree... Three Way Race with the NDP slightly ahead (within the MOE) LPC and CPC effectively tied.

[octothorp]

Originally Posted by Bownesian

Yeah, I can't explain Forum without dipping into conspiracy territory. How they could be as chronically off as they have been and not change their methodology is beyond explanation.

Wouldn't it be bad policy for a polling firm to see that they are an outlier and change their methodology to try to bring themselves closer to the mean?

I think that there may be problems with their methodology (based on their consistent outlier status), but polling firms should do their best to determine a good methodology, and then stick with it through the campaign, rather than adjusting their methodology to try and make their polls look as plausible as possible.

[Bownesian]

I agree totally.

I also remember the "Forum always polls the conservatives low" comments from what, three? elections ago so one wonders if their bias is a bug that they can't shake or (venturing into conspiracy now) a feature in that it produces answers closer to the client's desired response thus encouraging that voting behavior.

Either way, their clients continuing to hire them is tough to justify.

[Parallex]

Ipsos: http://www.ipsos-na.com/news-polls/p...e.aspx?id=6987

The Skinny: Same story... all parties within the MOE of each other.

[polak]

Originally Posted by PeteMoss

I wonder if this is mainly an Alberta thing (not sure where you actually live). The only time I've ever heard of strategic voting around here is against the Conservatives.

Also seems weird in the sense that at least budget wise the NDP are more aligned with the Conservatives than the Liberals.

I think when it comes to federal politics, economic concerns aren't nearly as far ahead in everyones mind as in provincial elections. Personally, social policy changes on a federal level are every single bit as important as economic ones when it comes to who is Prime Minister. That, plus the record for economic progress for the Cons and Libs isn't far off (personally, just reading some statistics lately, it looks like Liberals are better in that regard), it puts even more weight onto social views. Basically you have:

a) Liberal, progressive social agenda + decent economic record (ish)
b) Conservative, right leaning social agenda + decent economic record (ish)

[Parallex]

Riding Poll (Calgary-Confederation): http://www.mainstreetresearch.ca/lib...confederation/

[Parallex]

http://www.ctvnews.ca/mobile/politic...race-1.2571250

LPC: 30.8
CPC: 30.4
NDP: 28.9

[CorsiHockeyLeague]

Originally Posted by Parallex

Riding Poll (Calgary-Confederation): http://www.mainstreetresearch.ca/lib...confederation/

So on the face of it, you have a 33-32 lead (i.e. a functional tie what with MOE) and 14% undecided voters - which column would include me.

They note that the "leaning undecided" are leaning CPC... but it gets more interesting with the breakdown, though as it suggests that Matt Grant may have softer numbers... he has a 1 point lead in "certain to vote" respondents (again, functional tie) - but Webber has him by 13 points on "likely" voters. As for how strong the support for the party is, the CPC people are at 79% to 53% "strong" support, with 36% Liberals saying they "might change" versus 19% for the CPC, and 11% "undecided" liberal voters vs. 4% CPC. So I'm not sure how illusory the statistical tie is.

Obviously, the 14% of us who haven't settled are what matters, but I think if I were to bet on who is going to win just based on that I'd have to say Webber.

[Iowa_Flames_Fan]

Originally Posted by CorsiHockeyLeague

So on the face of it, you have a 33-32 lead (i.e. a functional tie what with MOE) and 14% undecided voters - which column would include me.

They note that the "leaning undecided" are leaning CPC... but it gets more interesting with the breakdown, though as it suggests that Matt Grant may have softer numbers... he has a 1 point lead in "certain to vote" respondents (again, functional tie) - but Webber has him by 13 points on "likely" voters. As for how strong the support for the party is, the CPC people are at 79% to 53% "strong" support, with 36% Liberals saying they "might change" versus 19% for the CPC, and 11% "undecided" liberal voters vs. 4% CPC. So I'm not sure how illusory the statistical tie is.

Obviously, the 14% of us who haven't settled are what matters, but I think if I were to bet on who is going to win just based on that I'd have to say Webber.

It's confusingly drafted, but the way I read the narrative is that 18% of the undecideds are "leaning" conservative, not that the undecided voters have a particular predominant slant.

How to allocate undecided voters is a bit of a complex issue, and usually involves making assumptions about what "late deciders" tend to do.

But I would not conclude from the smaller sub-sample of undecided voters that they are likelier as a group to vote conservative. Basically, of that 14% group, 18% are leaning conservative, and 13% are leaning either Liberal or NDP. Given the larger MOE of that subsample, that is pretty much what you are calling "a functional tie".

Not to mention that two-thirds of that 14% weren't leaning one way or the other at this point.

[CorsiHockeyLeague]

Originally Posted by Iowa_Flames_Fan

It's confusingly drafted, but the way I read the narrative is that 18% of the undecideds are "leaning" conservative, not that the undecided voters have a particular predominant slant. But I would not conclude from the smaller sub-sample of undecided voters that they are likelier as a group to vote conservative. Basically, of that 14% group, 18% are leaning conservative, and 13% are leaning either Liberal or NDP. Given the larger MOE of that subsample, that is pretty much what you are calling "a functional tie".

Their view is: "The advantage for Conservative Len Webber lies in the undecided voters who are leaning, where he leads with 18% to just 7% for the NDP and 6% for the Liberals." I'm not sure what's gained by your combining the Liberal and NDP leaning undecided numbers to get 13%?

Not to mention that two-thirds of that 14% weren't leaning one way or the other at this point.

Certainly true and as I say, early days so all of this will probably shift dramatically.

[FlamesAddiction]

"Undecided" is such a cop out way to answer a poll. They should simply ask that if the election was today and they had to decide (or just not vote), who would they vote for. Most reasonable people are still subject to changing their mind (or should be) within the next 4 weeks depending on what information is presented to them.

[Iowa_Flames_Fan]

Originally Posted by CorsiHockeyLeague

Their view is: "The advantage for Conservative Len Webber lies in the undecided voters who are leaning, where he leads with 18% to just 7% for the NDP and 6% for the Liberals." I'm not sure what's gained by your combining the Liberal and NDP leaning undecided numbers to get 13%?

Well, nothing--and that was actually my point. The "undecided Webber leaners" were less than 3% of the total sample, and only led "undecided leaners" for other parties by 1%. They are, in both cases, in a statistical tie with zero. My point was simply that no conclusions should be drawn from that about what undecided voters are likely to do a month from now.

[c.t.ner]

Originally Posted by Parallex

http://www.ctvnews.ca/mobile/politic...race-1.2571250

LPC: 30.8
CPC: 30.4
NDP: 28.9

I think this is probably the most interesting part of the polling data:

Among those who favoured the Conservatives, 45 per cent said they have no second choice, while 32 per cent said they would pick Liberals as their second choice, and 14 per cent said they would pick the NDP as their second choice.

Basically the Conservatives have a very strong lock on their base, which we all know. But if there is a shift over the next few weeks of Red Tories/Blue Liberals it'll to be the Liberals rather than the NDP.

[GullFoss]

Originally Posted by rubecube

We've been talking about strategic voting and what exactly is the path to victory for the Liberals, so I did a quick perusing of 308's riding projections, which are admittedly not completely accurate and I identified about 22 seats where either the NDP or Conservatives were leading by a slim margin and the Liberals were a close enough second that if 1/4 of the CPC voters switched to the Liberals, they would then win that riding. Based on the overall seat projections, the Liberals would only need to win half of those ridings to form a minority government, but they really don't have any path to a majority government.

Regionally they breakdown like this:

Ridings where the NDP are leading, Liberals close second:

BC - 4
AB - 1
MB - 1
ON - 1
QC - 1
NB - 1
NS - 1

Total: 10

Ridings where the CPC are leading, Liberals close second:

BC - 2
AB - 1
MB - 1
ON - 7
NB - 1

Thats true if there's only small swings in preference...but we also know it doesn't take much to tip everything over. Change begets change.

The Liberal path to majority is as follows:
1) The NDP started failing in the polls and conservatives started looking strong...causing mass vote switching to the Liberals (unite the centre-left) from the NDP

or

2) Conservatives started losing support causing mass vote switching to the Liberals from the conservatives (unite the centre-right)


What happened last election was a surging NDP and weakening Liberal caused mass vote switching to the Conservatives (unite the centre-right) giving them a path to majority. We can see that Liberals have gone back to their party now...hence the seat loss for the Conservatives and seat gains for the liberals.