Latest polling thread July 14, 2015

Norman Spector ‏@nspector4  4h4 hours ago

Norman Spector retweeted Globe Politics

Does @paulisci understand that the #NDP are far ahead in seat-rich francophone areas of Quebec?

Norman Spector added,

Globe Politics @globepoliticsQuebec the most volatile electoral battleground: election simulator http://bit.ly/1OohPQL  #cdnpoli

nicky wrote:

http://www.guelphmercury.com/news-story/5705222-liberals-can-hold-on-to-... This article refers to a Forum poll from about a month ago showing the NDP ahead in Guelph. 34 to 28 for the Liberals and 26 for the Cons. This would be a remarkable gain for the NDP and perhaps reflects why Mulcair is campaigning heavily in SW Ontario. No reference in story to the so- called star Green candidate Gordon Millet but he cannot be polling more than 12% and maybe less. I have tried to find further details of this poll without success.

Oh yes. Despite all the griping about Andrea Horwath, the Ontario election did a lot to institutionalize the NDP in South-Western Ontario.

It is also not good for Mulcair to be discussing coalitions either at this point. It sends a message to voters that he believes that the NDP is only minority government material. He should be talking majority and thinking exclusively majority. It gives voters a better sense of confidence that Mulcair is the one to vote for. Set your goals high and show your confidence as a leader and leave the coalitions to discuss after the election if that is what the results come to.

Misfit wrote:

It is also not good for Mulcair to be discussing coalitions either at this point. It sends a message to voters that he believes that the NDP is only minority government material. He should be talking majority and thinking exclusively majority. It gives voters a better sense of confidence that Mulcair is the one to vote for. Set your goals high and show your confidence as a leader and leave the coalitions to discuss after the election if that is what the results come to.

NDP has campaigned successfully twice by remaining open to coalition once in 2011, and all through 2014/15 with the polling result we see now. Its tempting to try and dry out the Liberals using a strategic voting stretcher, but NDP positive openness in contrast to Iggy's negative closure helped set the tone for the positive campaign that created the Orange Crush.

It would be an utterly transparent vote buy for Mulcair to flip-flop on this, when many have questioned previous statement against "cooperation" in the NDP leadership campaign --- there and back again?

As long as Trudeau doubles down the "no coalition" line, the contrast will remain, and he may have said "no" too often, and too definitively to back track now.

Liberal strategy at this point seems to be damage control, and holding onto the base, hence the obnoxious "Mulcair is a separatist" attack ad, recently released -- this is of no interest to anyone but LPC veterans of the sovereignty wars of the 80s and 90s. Remember, success for the third place Liberals will be judged at a bear minimum of doubling their seat count, with "success" measured as becoming official opposition -- that will be their target now.

Winning or bust is not a precondition for Trudeau, as it is for Mulcair.

Staying on track with positive "coalition" talk will also reduce much post-election political fuss of "legitimacy" should the NDP try a coalition with LPC, since they are campaigning on it.

Unseating Harper is what those tempted by "strategic voting" want so there is probably as much to be gained from defining the NDP around being willing to unseat Stephen Harper one way or the other, as soon as possible after the election, and Trudeau rejection of that will sour his image with that demographic.

Rokossovsky wrote:

Misfit wrote:

It is also not good for Mulcair to be discussing coalitions either at this point. It sends a message to voters that he believes that the NDP is only minority government material. He should be talking majority and thinking exclusively majority. It gives voters a better sense of confidence that Mulcair is the one to vote for. Set your goals high and show your confidence as a leader and leave the coalitions to discuss after the election if that is what the results come to.

NDP has campaigned successfully twice by remaining open to coalition once in 2011, and all through 2014/15 with the polling result we see now. Its tempting to try and dry out the Liberals using a strategic voting stretcher, but NDP positive openness in contrast to Iggy's negative closure helped set the tone for the positive campaign that created the Orange Crush.

It would be an utterly transparent vote buy for Mulcair to flip-flop on this, when many have questioned previous statement against "cooperation" in the NDP leadership campaign --- there and back again?

As long as Trudeau doubles down the "no coalition" line, the contrast will remain, and he may have said "no" too often, and too definitively to back track now.

Liberal strategy at this point seems to be damage control, and holding onto the base, hence the obnoxious "Mulcair is a separatist" attack ad, recently released -- this is of no interest to anyone but LPC veterans of the sovereignty wars of the 80s and 90s. Remember, success for the third place Liberals will be judged at a bear minimum of doubling their seat count, with "success" measured as becoming official opposition -- that will be their target now.

Winning or bust is not a precondition for Trudeau, as it is for Mulcair.

Staying on track with positive "coalition" talk will also reduce much post-election political fuss of "legitimacy" should the NDP try a coalition with LPC, since they are campaigning on it.

Unseating Harper is what those tempted by "strategic voting" want so there is probably as much to be gained from defining the NDP around being willing to unseat Stephen Harper one way or the other, as soon as possible after the election, and Trudeau rejection of that will sour his image with that demographic.

Full agreement. 

Forum: NDP 34 (134 seats) Liberals 29 (78 seats) Con 28 (121) Bloc 4 seats, Greens 1.

http://poll.forumresearch.com/data/Federal%20Horserace%20News%20Release%20(2015%2007%2021)%20Forum%20Research.pdf

Conservative support among voters jumps as child care cheques land on doorsteps poll finds

http://news.nationalpost.com/news/canada/canadian-politics/conservative-...

Did polling in the Alberta race. Don't think they have much, if any, of a national track record. Outlier.

They were polling as the cheques arrived in the mail

https://en.wikipedia.org/wiki/Alberta_general_election,_2015#Opinion_polls

 Because the numbers in Ontario are bullshit, no one else sees them or anything close, forum came out with at poll at the same time period and it doesn't show it.

 Look at for who the poll was designed and how.

I have not been on rabble this week but I wanted to come and share this:

I raised a number of questions about this Mainstreet poll on Twitter today. The numbers don't add up. While I do not question their data or methodology in terms of collection, I suspect that when they prepared the data for release numbers were transposed in error.

Here is the link to the release (the PDF is at the bottom but not the excel sheet of course): http://www.mainstreetresearch.ca/2015/07/24/conservative-christmas-in-july/

In their release they say those who are following the release of the UCCB cheques closely AND are eligible to receive them, are much more likely to vote Conservative (heading to para 4). This subgroup is reported at 38% for the CPC.

They also say that those who are not aware of the UCCB cheques are much less likely to vote CPC (23% -- but reported as 25% in the NP).

(There have been reports in the media during the last week stating that those who are eligible to get the UCCB cheques are more likely to be Conservative and more likely to be in Conservative ridings.)

Then they say that the national figures for Conservative popularity is 38% -- the very same number (para 3) as the subgroup of those getting cheques.

Even if the CPC had increased for other reasons across the board you would expect support from those receiving cheques to go up in tandem. How would it make sense that the rest of the population could catch up while they remained the same? It is difficult to believe that a group identified as core support for the CPC would have identical support to the national result.

It seemed more likely that the figures provided for the national poll has been overwritten in the data excel by the subgroup figures OR that they took the figures for the subgroup as the national figures.

That's a pretty coherent explanation. It would be odd for the subset most likely to support the CPC to be the same as the overall poll. Something is defintely wrong. 

"Quito Maggi, president of Mainstreet Research."

Also known for pieces like this

http://www.huffingtonpost.ca/quito-maggi/rob-ford-2014-campaign_b_5045394.html

and this

http://www.huffingtonpost.ca/quito-maggi/ford-nation_b_5878820.html

I did an analysis of the available information about the demographics of the Mainstreet Poll. Very little information is available but what we do see does not suggest reason for confidence. Since I did this work, I may as well post it here.

I approached them and they offered me tables to see the math. Then they sent me the UCCB eligible table 1044 respondents and then said I could not see the rest as it was private between them and the client. The client is a newspaper and the poll is a publicly released poll.

You can see the demographic breakdown if you look at their poll and you can see some significant distortions.

They interviewed 5147 people as follows:

# ages %

534 18-34 10%

860 35-49 16.7%

1698 50-64 32.9%

2055 65+ 39.9%

As you can see only 10% of their survey were people under 34. Elections Canada has provided an estimate of the voting demographics here: http://www.elections.ca/res/rec/part/estim/41st_GE_turnout_e.xls

18-24 7.7%

25-34 12.7%

35-44 15.65%

45-54 22.0%

55-64 20.37%

65-74 12.9%

over 75 8.5%

These demographics do not overlap in all age ranges but some do. Let's compare Mainstreet age demographics to actual voting statistics.

18-34 Mainstreet weight 10% Elections Canada 20.4% Mainstreet under estimate by half or 10%

65+ Mainstreet weight 40% Elections Canada 21.4% Mainstreet over estimate by almost double or 19.6%

My estimates as Mainstreet is 35-49 and 50-64 and Elections Canada is 35-44, 45-54 and 55-64.

35-49 Mainstreet 16.7% Elections Canada 15.65 and portion of 22 let's take 10% = 25.65% Mainstreet underestimate by 1/3 or 9%

50-64 Mainstreet 32.9% Elections Canada 20.37 and portion of 22 let's take 12% = 32.37% Mainstreet seems to have this correct.

This age distortion is so significant that the survey would be very compromised.

We have no idea what other distortions exist but we have a couple clues as to more:

They surveyed 1044 parents of kids under 18. they only surveyed 1394 people under age 49. While some parents may be over 50 the vast majority would be expected to be under 50. This means that the number of young people they surveyed were tilted heavily to families with children and very few single young people were interviewed.

Now lets look to see how they did weighting provinces

In 2011 here are the following percentages of total votes from each province (http://www.elections.ca/scripts/ovr2011/default.html)

Atlantic 7.7% 8.1%

Quebec 25.8% 27.0%

Ontario 37.5% 31.16%

Prairies 6.4% 8%

Alberta 9.4% 9.8%

BC 12.7% 15.8%

Overall much closer but still a few differences.

The Mainstreet survey gender split was 45% male and 55% female.

Women do vote slightly more than men: women participated at a higher rate (59.6%) than men (57.3%) and they represent a higher share of the population but the difference is only 1% of the population (49.5 male 50.5 female). The difference Mainstreet has is much greater. I have to assume this was due to distortion of younger adults being more likely to have children (more families are headed by women than men).

So what we see is a significant mystery when it comes to much of the methodology but some serious questions about the methodology we can see.

All this is a case in point for one survey showing you how demographics can distort the result.

Of course you can also distort a result more by taking more sample from rural areas (easier and cheaper to survey than cities if by phone), particular ridings, specific ridings and poll districts etc. And then there is the question of how the survey itself is conducted-- the order of questions etc.

There is a lot of reason not to trust polls in general. This particular howler provides a case study of the types of problems and the lack of accountability so the public cannot verify all that much.

I am not following these threads very much right now  -- if you want to ask me a question about this survey please send me a DM -- I will geta notice in my email and come and get any messages. Otherwise I may miss a reply.

Thanks for this analysis, Sean.

To sum on the age demographics for Mainstreet:

If a survey were weighted to reflect actual percentages of voters in each age range (according to Elections Canada) the number of people surveyed under 34 should be close to the same as the number over 65. However, Mainstreet has four times as many over 65 as under 34. A 400% difference. Of course that means something.

bekayne wrote:

Sean in Ottawa wrote:

To sum on the age demographics for Mainstreet:

If a survey were weighted to reflect actual percentages of voters in each age range (according to Elections Canada) the number of people surveyed under 34 should be close to the same as the number over 65. However, Mainstreet has four times as many over 65 as under 34. A 400% difference. Of course that means something.

They say they weighted by age, though they don't mention the census, as other pollsters do.

The poll surveyed a random sample of 5,147 Canadians by Smart IVR™ on July 20-21, 2015.Mainstreet surveyed a mixture of landlines and cell phones. Results were weighted by age,gender, region and language.

A survey is weighted by the quotas you set for each demographic as you run it in field. This is done so you do not do piles of surveys you don't need to do. Age qualifiers are asked at the start of the survey -- may I speak to a person between the ages of x and y. As you do the survey and fill quotas x and y are modified.

You can see in their PDF that they are issuing both percentages and numbers for all 5147 respondents-- and the age ranges. The quota in each category they use is the weight. We know how many of each age they have as they published that. And they claim to have the weight of 5147 surveys when they claim their accuracy.

Looking at the demographics it looks like they did close quotas somewhat in regions but not appropriately in terms of age (if at all) -- unless they mean they "weighted by age to distort the survey in order to have the Conservatives lead..." Possible but I doubt it. More likely they did what was cheapest to get a big result at the least cost of sample.

Gender: they used 10% more women than men (I suspect this was to make it easier to get qualified parents as there are more female headed families than male).

Region: the provinces are semi-accurate -- I did analyse this and the results are off but not drastically.

Age: the weighting for age is what I have provided -- it is in their results. They are using results where 72% of the surveys come from people over 50 and 40% from people over 65. 10% for people under 34.

This IS their weighting. If they discarded any over-quota results above that we don't see this.

And let's be clear they are claiming the data is accurate becuase they used 5147 respondents. If they were to weight this then that number would have to go down to represent the number of surveys used in that particular line. Since this 5147 includes 4x the seniors as under 34 year olds when they should be equal, the actual size of survey where you could get balanced data would drop by 75% so we would no longer be speaking of a poll of 5147 but rather under 1500.

Either they are lying when they say that the topline party preference represents 5147 people or they are not being truthful when they say it is weighted. Of course they did not say it is accurately weighted or weighted to match voters or the population.

Canada is aging but Seniors do not yet represent 4 in 10 of us. And over 50 does not represnt more than 7 in 10.

It looks clear they used the entire data set. when you look at the PDF.

Pierre C yr wrote:

Are they weighing for likely voters? People over 65 vote at a far greater ratio than those under 34.

This would not make sense. The data I used is NOT population data which you would then have to calculate who are citizens and then apply an estimate re likely voters. This is already actual VOTER data from Elections Canada. The fact that older people vote more than younger people is built into this Elections Canada document becuase this is the actual voters. Elections Canada in the same chart provided eligible voters for each age group but I ignored that and took only numbers of voters for each age group. There is no reason to presume that the dynamic of likely voters would change that much in the last four years.

Just so you know the numbers the actual voting participation rate goes from 38% in 18-24 range to 75% in the 50-64 range (it declines after that). But as I say this participation rate is already built in to the numbers I used. BTW Canada has 34 million residents. It has only about 14 million voters. This is after you consider those who are too young and those not eligible like permanent residents.

The elections Canada statistic is the most accurate statistic to work from as it does not guess which ages might vote -- it provides data on who actually does vote in each range. They know who votes and who does not from voter lists and they know the ages of voters as this is recorded when you vote if it is not included in the tax derived voter list.

Now the one thing you might consider is population aging. However, population aging is very minimal over 4 years in these ranges. As well the differences would be more within the ranges rather than between them. For example:

1) some will go from under 18 to over 18; some from under 35 to over 35. These two cohorts are not much different as there is a boomer echo that is turning 18.

2) The group moving from Age 46 to 50 would make little difference either as they are offset by a similar cohort moving from 61 to 65.

3) The cohort moving past 65 is not completely offset by people dying off but the difference is not massive over 4 years (it is more significant over a longer period). So for example, the other cohorts would remain just about the same but the over 65 group might go from the current 21.4% to perhaps 22.5 or 23% But it sure won't explode to 40% as Mainstreet has it.

So, no -- there is no need to weigh the figures I am using any more than they are unless you are trying to distort results...

http://www.theglobeandmail.com/news/politics/mulcair-leads-election-poll...

nicky wrote:

http://www.theglobeandmail.com/news/politics/mulcair-leads-election-poll...

It's great news.  People are really warming up to the NDP.  I'm looking forward to the Ekos poll.

One interesting quote from the article was this:

Quote:

Some also said the federal Conservatives, which have focused their efforts on hammering Mr. Trudeau, have yet to turn their firepower on Mr. Mulcair and the NDP. “Is Mulcair looking better having nobody taking a critical look at him?” one Conservative strategist said. “I’m sure that is helpful for him at the moment. I suspect that will not be the case for the next three months.”

For the summer I suspect they'll keep attacking Trudeau, since there are still votes for the Cons in Ontario that can be dislodged from the Libs.  But by September, if things still appear to be in the NDP's favour, then I suspect that the Cons will focus on the NDP.  Possibly sooner, but that would depend upon the polls.

double post sorry 

FiveThirtyEight's editor-in-chief (and America's most successful political prognosticator) is actually the only one in North America worth following. All the rest are either incompetent or are connected to a political party.

Nate Silver thinks Bernie Sanders could beat Hillary Clinton in Iowa and New Hampshire

Before you write off a double Sanders victory as unlikely, remember that in the 2008 general election, Silver called 49 out of 50 states correctly, and in 2012, he nailed all 50. Ben Frumin

Brachina wrote:

 This is disasterous news for the Tories, the economy is the.issue they tradionally dominate, to be losing it to Tom Mulcair and the NDP, its a huge crisis.

 It'd be like the NDP losing the issue of healthcare to the Tories or the Greens losing the enviroment to the Libertarians.

I thought the Greens were Libertarians