New poll: Harper - Conservatives headed for third place next election?

[burn_this_city]

http://www.ctvnews.ca/politics/trude...psas-1.1961682

I guess we're going to get inundated with Health Canada ads related to marijuana. I see the conservatives losing a degree of support over their stance on pot.

[Sylvanfan]

Originally Posted by killer_carlson

Edmonton already has 1 seat NDP (Strathcona, a traditional NDP provincial seat). It would not surprise me to see the NDP take another in Edmonton, and the PC seat held by Laurie Hawn go back to the Libs (who used to have Anne McClellan in the riding).

PCs are not slam dunks in every riding.

Ratheberger as an Independent in St.Abert will not be a push over either. As the incumbent MP who quit the Conservatives he gained some popularity with voters.

[Iowa_Flames_Fan]

Yeah I'll admit I'm not as familiar with the conditions on the ground in Edmonton. But if the Tories lose 4 seats in Alberta, then the Liberals will win a majority government, I think.

It would be interesting to post some predictions given the unusually stable polling over the past number of months. The numbers right now look really bad for the Tories, but there are of course two very important caveats, both of which have been raised already in this thread: one is that we are still very far from an election (and it's unlikely there will be an early election call given that Harper is dipping in the polls), and a lot can change even after the writ is dropped.

The other is that Canada is hard to poll, for reasons that I don't think are well understood. My belief is that it has to do with the fact that a federal election is not one election but 308
separate elections, each of which has a separate, relatively tiny electorate. This affects not only voter intention, but voter behaviours; a Liberal in Red Deer may tell a pollster that he intends to vote Liberal, but decide not to bother on Election Day because his candidate is going to be annihilated.

I'll add a third, which is the in-this-case aptly named "shy Tory
effect." In essence, where an incumbent is embroiled in scandal and so personally unpopular that it is embarrassing to admit that you intend to vote for him anyway, voters will tell pollsters that they are switching their votes only to have a last minute change of heart in the booth. In this case, Harper has very publically bungled a few things (his spat with the Chief Justice is just one example) and this might lead to an "illusory" dip in his polling numbers that we might afterward attribute to "shy Tories".

All of this is mere speculation, but that to me is part of the fun of poll watching. If I had to predict I would say that Trudeau is headed for a minority government, but I wouldn't dream of putting any money on it.

[troutman]

Shy Tories - good name for a band.

[Rerun]

Originally Posted by Iowa_Flames_Fan

Yeah I'll admit I'm not as familiar with the conditions on the ground in Edmonton. But if the Tories lose 4 seats in Alberta, then the Liberals will win a majority government, I think.

It would be interesting to post some predictions given the unusually stable polling over the past number of months. The numbers right now look really bad for the Tories, but there are of course two very important caveats, both of which have been raised already in this thread: one is that we are still very far from an election (and it's unlikely there will be an early election call given that Harper is dipping in the polls), and a lot can change even after the writ is dropped.

The other is that Canada is hard to poll, for reasons that I don't think are well understood. My belief is that it has to do with the fact that a federal election is not one election but 308
separate elections, each of which has a separate, relatively tiny electorate. This affects not only voter intention, but voter behaviours; a Liberal in Red Deer may tell a pollster that he intends to vote Liberal, but decide not to bother on Election Day because his candidate is going to be annihilated.

I'll add a third, which is the in-this-case aptly named "shy Tory
effect." In essence, where an incumbent is embroiled in scandal and so personally unpopular that it is embarrassing to admit that you intend to vote for him anyway, voters will tell pollsters that they are switching their votes only to have a last minute change of heart in the booth. In this case, Harper has very publically bungled a few things (his spat with the Chief Justice is just one example) and this might lead to an "illusory" dip in his polling numbers that we might afterward attribute to "shy Tories".

All of this is mere speculation, but that to me is part of the fun of poll watching. If I had to predict I would say that Trudeau is headed for a minority government, but I wouldn't dream of putting any money on it.

Also, could it be that the ones who most tend to participate in these polls are the "squeaky wheels"?

[MarchHare]

Originally Posted by Rerun

Also, could it be that the ones who most tend to participate in these polls are the "squeaky wheels"?

That's not how polling works.

[GP_Matt]

Don't forget the fact that the government is tracking for a strong surplus next year.
Billions of dollars can buy a lot of votes. Whether it is targeted tax breaks or spending announcements they are in a position to really let the money fly.

The other unknown is the European free trade agreement. The text of the document should be out before the election and that could really shift peoples vote depending on the reaction.

[EldrickOnIce]

Originally Posted by Bobblehead

So?
Then one of their superstar candidates started discussing a referendum and separation.

There is no way an early poll can account for that, or candidates offering racist views (WR in Alberta) or a leader announcing he is going to cut 100K jobs (PC in Ontario). Polls may be valid when they were taken but then something happens.

You answered your own dickish question.
See above...

Polls are not necessarily predictive, simple as that

[Rerun]

Originally Posted by MarchHare

That's not how polling works.

Really? Someone calls me up and asks me to participate in a poll... but I'm just an average Joe and really not interested in politics and I figure I've got better things to do than answer a bunch of silly questions.... so I say no and hang up. The pollster calls the next number and gets someone who is dissatisfied with the way things are being run.. so he says yes because he wants to get a chance to put his 2 cents worth in and perhaps change the way things are being done by increasing the % of voter dissatisfaction?

[MarchHare]

Originally Posted by Rerun

Really? Someone calls me up and asks me to participate in a poll... but I'm just an average Joe and really not interested in politics and I figure I've got better things to do than answer a bunch of silly questions.... so I say no and hang up. The pollster calls the next number and gets someone who is dissatisfied with the way things are being run.. so he says yes because he wants to get a chance to put his 2 cents worth in and perhaps change the way things are being done by increasing the % of voter dissatisfaction?

With a statistically significant sample size, effects like you described are eliminated.

[Iowa_Flames_Fan]

Originally Posted by Rerun

Also, could it be that the ones who most tend to participate in these polls are the "squeaky wheels"?

It's possible; I'm not sure of the dynamics of who responds and doesn't respond to polls in Canada. However, my guess (and it's only a guess) is that people who respond to polls would usually also be likelier than the average person to actually vote.

[SebC]

Originally Posted by MarchHare

With a statistically significant sample size, effects like you described are eliminated.

Sample size reduces random fluctuations, not systematic errors.

[Devils'Advocate]

Originally Posted by MarchHare

With a statistically significant sample size, effects like you described are eliminated.

Not true. Not true AT ALL. I am a system designer for Statistics Canada. When you are first hired into a professional position at STC they send you out on a course where you are put in the field to get to know what the agency does. So even though I work in IT, I was put on the front lines knocking on doors.

First we picked the houses randomly. We then got a list of the people in the house and chose randomly. If that person was not at home we had to come back on 4 separate occasions in order to get the person we selected. We asked the instructor why we had to be so specific - why not just ask the person who answers the door. We were told that if you just gets the person that answers the door you are going to oversample retired people, unemployed people, housewives, etc.. and under sample people that are usually at work, busy parents always off with their kids and people who refuse to answer their door.

On our phone interviews, we will call a random person, and if you say you are too busy we will ask for a time that is more convenient. If you say that you don't know when you will be available, we will wait a few days, maybe a week and call you back. For us it is VERY important that we actually interview the people that we randomly sample because if you don't you introduce bias into your survey. Very few polling companies are as persistent as we are, so there is a lot of inaccuracy in their numbers. They are not getting a true random sampling of the population.

BTW - I was the "dog person". The person that got called upon to knock on the door of the houses that had "Beware of Dog" signs. This wasn't because the rest of team didn't like me, but rather because I am really, really good with dogs.

[MarchHare]

That's interesting stuff, DA.

Do you agree with Rerun, then, that the polling firms are over-sampling people dissatisfied with the present government and therefore inflating Liberal/NDP support?

[Devils'Advocate]

I try not to agree with Rerun.

All I am saying is that unless a polling company is scrupulous in their methodology, there will be biases and not a true random sampling. I do know that almost all polling companies oversample older people and under sample parents of school aged children and young people. Why? Because older people are the most likely to answer a random call out of the phone book. Younger people either don't have a number in the phone book or they are not at home to answer it. So knowing that there are biases, some of which can be identified and some we don't even think of, we have to take polls with a grain of salt. They can show trends and can give ballpark numbers, but don't expect exact predictions. If they were to follow our methodology, they would be too expensive to run. Rerun *could* be right. He might not be. But sampling size increases would only get you MORE elderly people with land lines. When your sample is not truly random, no matter what your sample size, biases of all kind can take place.

[Iowa_Flames_Fan]

Originally Posted by Devils'Advocate

I try not to agree with Rerun.

All I am saying is that unless a polling company is scrupulous in their methodology, there will be biases and not a true random sampling. I do know that almost all polling companies oversample older people and under sample parents of school aged children and young people. Why? Because older people are the most likely to answer a random call out of the phone book. Younger people either don't have a number in the phone book or they are not at home to answer it. So knowing that there are biases, some of which can be identified and some we don't even think of, we have to take polls with a grain of salt. They can show trends and can give ballpark numbers, but don't expect exact predictions. If they were to follow our methodology, they would be too expensive to run. Rerun *could* be right. He might not be. But sampling size increases would only get you MORE elderly people with land lines. When your sample is not truly random, no matter what your sample size, biases of all kind can take place.

This is a great post. I'd only add that the sampling errors you've described would be unlikely to import an anti-Conservative bias into the polls, and demographically would more probably lead to over sampling of groups that are historically likelier to be Harper voters.

With that said, there may well be other sources of error at issue here.

[SebC]

Originally Posted by Iowa_Flames_Fan

I'd only add that the sampling errors you've described would be unlikely to import an anti-Conservative bias into the polls, and demographically would more probably lead to over sampling of groups that are historically likelier to be Harper voters.

This is almost certainly the case. People who own landlines are generally more conservative.

[Jacks]

Originally Posted by SebC

This is almost certainly the case. People who own landlines are generally more conservative.

Any proof of that?

[PsYcNeT]

Originally Posted by Jacks

Any proof of that?

Old people are generally more conservative.

http://www.ropercenter.uconn.edu/ele.../voted_12.html

And old people are more likely to have a landline.

http://gigaom2.files.wordpress.com/2...less201212.pdf (page 8)

[Mike F]

Talking about statistical theory and polling strategy effectiveness is all fine, but here's some real-world results:

Last federal election, the results for the 3 major parties by % were:
CON - 39.6%
NDP - 30.6%
LIB - 18.9%

Now, here's a summary of polling around the 2011 election. In the immediately preceding days, the polls were a couple points off, but close. You can decide for yourself whether that is close enough to consider modern polling, even with its inherent biases, accurate.

However, you can also see a long term polling summary from 2008 to 2011 here that shows the inherent danger of reading anything into polls this for from ballot day.