Polling thread infinite edition: Volume 7

102 posts / 0 new
Last post
remind remind's picture
Polling thread infinite edition: Volume 7

continued from here

NorthReport

bump

ottawaobserver

OK, I'm going to try this one again with the polling methodology gurus (I exclude myself from that group, for sure):

Sean in Ottawa says it's not the prompting, but EKOS has the Greens higher than anyone else.  Sean also says that Greens self-select into surveys more, but that doesn't explain the high (I think we all agree it's too high) Green number.

So, what else could explain that (I think we all agree it's HUGE) gap between Nanos and EKOS?  They can't both be right.

Is it the technology?  It is that some polling firms prompt with the leaders' names attached to the parties, while others just prompt with the party names?  Is it the sampling technique?  Is it the introduction of the poll as being *about* politics.

I know Sean has written at length in general about these factors, but never specifically in relation to explaining the specific difference between Nanos and EKOS.

Sorry if I'm being a bit thick here, but I'd like to benefit from your perspective.

Also, do we have any idea why EKOS starting adding "Other" to the options they're prompting with?  Could it be they're trying to establish a baseline for "Other" in order to try a split sample (even if Sean doesn't think that's a problem)?

KenS

Good questions.

Doug
Sean in Ottawa

ottawaobserver wrote:

OK, I'm going to try this one again with the polling methodology gurus (I exclude myself from that group, for sure):

Sean in Ottawa says it's not the prompting, but EKOS has the Greens higher than anyone else.  Sean also says that Greens self-select into surveys more, but that doesn't explain the high (I think we all agree it's too high) Green number.

So, what else could explain that (I think we all agree it's HUGE) gap between Nanos and EKOS?  They can't both be right.

Is it the technology?  It is that some polling firms prompt with the leaders' names attached to the parties, while others just prompt with the party names?  Is it the sampling technique?  Is it the introduction of the poll as being *about* politics.

I know Sean has written at length in general about these factors, but never specifically in relation to explaining the specific difference between Nanos and EKOS.

Sorry if I'm being a bit thick here, but I'd like to benefit from your perspective.

Also, do we have any idea why EKOS starting adding "Other" to the options they're prompting with?  Could it be they're trying to establish a baseline for "Other" in order to try a split sample (even if Sean doesn't think that's a problem)?

 

Nicky posted these numbers in the last thread:

Ekos 9.6

Angus Reid 7

Nanos 8.2

Harris 9

Strategic Counsel 11

Ipsos Reid 8

Segma 10

Firstly, Ekos is not the highest. Secondly I think they all now probe for the party names including the Greens. I checked several of them you can see they do and others the way they structure the poll implies they do. Still the Green vote is volatile from one survey to the next-- conducted using the same methodology by the same firm. This is more consistent with the issue being a small number self-selecting but very difficult to to predict than a large number that need to be coached or probed. Where you call makes a world of difference-- how you make up the Provincial numbers is based on a formula for how many calls in each demographic and locale. Each pollster has different methodologies driven by cost, bias and opinion. As I said before polling is an art not a science.

I never worked for either Nanos or Ekos. It is unrealistic to expect a specific analysis of those two firms in isolation using what would be confidential infromation related to their methodology. That said niether of those firms have been very consistant anyway when it comes to Green support. general trends can be found but it is more than probing-- there is not that significant a number of people who will agree to do a political poll but don't rememebr without a prompt who they support  (think about that for a moment and you can see how silly this probing thing really is as an explanation). Most who are not political simply won't answer the question-- if they are they will give a name.

One difference in probing may be the interviewers-- many are very low in skills and judgement. The interviewers if they do not have a box to put the responses in may simply not count the Green responses putting in DK, hanging up (believe it that happens), terminating otherwise, placing it randomly somewhere else. Don't forget about interviewer error-- I monitored enough calls to know this is a significant factor and having the complete list makes a difference. But still this is not enough to explain the variations--

Anyway I have answered the questions a couple times and there is no evidence and not even common sense supporting the theory that a 30-50% change can be explained simply by probing the name of a party. Again can you really imagine tons of people who don't know the party they would support, and it end up being all one party when prompted, and that these would be willing to do a survey in spite of their ignorance and lack of interest? Defies sense. This is a lazy theory by people who cannot come up with a more plausible explanation and are unwilling to accept that the numebrs are just not that accurate.

 

Sean in Ottawa
remind remind's picture

Personally, I think the numbers on choice surprised them, as they never released the numbers from the AR survey they took 2 weeks ago on it.

 

If it had been what the right wing whackos were looking for they would have been trumpeting it across the air waves.

 

 

 

Debater
Stockholm

Its not that surprising, back in the 70s and 80s, polls would routinely show that support for the death penalty in canada was about 75%. But it all depends on how you ask the question. Even referring to "capital punishment" as opposed to "the death penalty" makes a big difference. Also, if you ask people whether they favour the death penalty for murder period - there is no alternative offered. The results can be quite different when people have to choose between the death penalty for murder and life imprisonment with no possibility of parole for murder. 

ottawaobserver

Sean in Ottawa wrote:

This is a lazy theory by people who cannot come up with a more plausible explanation and are unwilling to accept that the numebrs are just not that accurate.

You know, I'm a lot of things, but I'm NOT lazy.  Sorry I bothered you.

Sean in Ottawa

Ottawa Observer-- I am sorry I did not mean it that way-- I have been getting tired of this being bandied around like there is some actual proof for it when there is not.

The theory only has repetition going for it-- not facts.

I did not mean to imply you were lazy-- I do think that the theory is an easy way to explain something that is a lot less simple and I am tired of being asked the same question I have already answered..

KenS

Hmmm...

While you may have answered the questions, I don't have that impression either.

Likely contributing factor: when you give long and multi-staged answers, you've got to expect that people don't absorb it all. [Some else does that too. :)] At a minimum- you have to expect to do what seems like repitition. And when you do, work on pruning the thing down in the process.

nicky

 I very much appreciate the efforts made to explain the discrepencies but I am still far from convinced about any of the theories advanced for the high Green support in some polls.

I still tend to favour the "prompting" explananation. The figures I posted earlier were the final Green predictions by the various pollsters before the last election. Sean makes the valid point that the EKOS prediction for the Greens was right in the middle, although still about 50% too high.

Perhaps the answer is in the lack of commitment by the typical Green voter rather than their zeal in participating in polling. I think the polls generally show that the Green voter is more likely to switch than supporters of other parties. To use an overused term, people "park" their votes with the Greens before they make their final choice.

An apathetic or genuinely uncommited voter might be more inclined to say he was uncommitted if unprompted by the pollster with a list of parties. If prompted such voters might well split more or less evenly among the choices offered. On a random basis the Greens would get about a quarter of these people rather than the 5-6% which might be closer to their true vote. Perhaps this is why  prompted polls overrepresent the Green vote.

I would be very intersted if someone has figures swhich show that particular pollsters conssitently register higher Green support. My figures were simply the final pre-elaction predictions. What is the trend of the various pollsters over an extended timeframe?

Sean in Ottawa

That's the point -- there is no trend it is all over the map -- look at poll trackers over time that include all the polls and you will see the Greens tend to be up and down for the same poll/methodology.

It is not that difficult a theory that the polls are simply less accurate than people imagine they are.

Sean in Ottawa

In fact all parties swing by about the same number-- it is just that the Greens being a small number gather more attention. So the swing is inaccuracy.

The higher than accurate range for the Greens I have already explained several times.

Another point is that the Greens do not have any ground support to identify and get out the vote-- other parties put a lot of effort in that.

I keep raising this because I am tired of the borderline insults to Green voters being trotted out as an explanation when there are better explanations that hold up-- perhaps those are less flashy and don't serve anyone's agenda.

Sean in Ottawa

Nicky why do you favour the prompting explanation -- just because it is mentionned more often? I just hear people arguing for it for default even though there is nothing to back it up other than empty assertions-- like those who probe for the Greens having higher numbers-- not supported.

KenS

Maybe people subsribe to the prompting explanation because they have not heard something better that thay can remember.

I'm open to alternative explanations. Don't really remember yours.

I gave some unmistakable hints in that direction already. Answering OO's questions would be one way to summarize, but there may be others.

Stockholm

I favour the prompting explanation because there is clear evidence to support it. Nanos is the only one who does not prompt any of the party names - and he almost always has Green support about half of what anyone else has (and he when the votes are counted even Nanos ends up overestimating them). Who could forget the farce of the November byelections where in the midst of a flurry of national polls claiming that the Greens might get as much as 12-13% of the vote they got a humiliating 2-3% even in places that should have been relative strongholds for them like a rural Nova Scotia seat right next door to Central Nova or a Vancouver riding.

KenS

I don't agree with that last point Stock. The GPC performs below even its own average in the type of riding represented in all 4 of those by-elections. May knows that going to Central Nova was a stupid move, they poll poorly here.

Augustus

The Conservatives have the ability to start rising in the polls if they start focusing on the economy.

 

The main issue that concerns Canadians right now is the economy, not prorogation.  Most Canadians will be paying attention to how they are going to fare economically, rather than on the prorogation rallies.

 

Harper is already beginning to focus on the economic direction of the country:

 

 

"We've obviously had a successful year rolling out infrastructure projects, but we now have to turn our mind to the broader agenda, to some of the economic challenges including deficit reduction ahead of us,"

http://www.businessweek.com/news/2010-01-23/harper-says-canada-governmen...

Sean in Ottawa

KenS wrote:

Maybe people subsribe to the prompting explanation because they have not heard something better that thay can remember.

I'm open to alternative explanations. Don't really remember yours.

I gave some unmistakable hints in that direction already. Answering OO's questions would be one way to summarize, but there may be others.

In fact I did-- in the last thread I wrote a lot about this.

Where and how they poll, and that only a very, very, very, very, very small minority of people are willing to answer a pollster and they are not representative of those who don't

As well there are lifestyle and geographic reasons-- like where and when you work and when you get home-- these polls tend to be done early evening-- people who do shift work or get home late or even those with young familes are less likely to participate in a poll. Pollsters likesmaller communities because they don't have the bulk of the population in their cars for 45 minutes during the best polling time.

The single methodological difference of to prompt or not does not explain why pollsters who prompt show wild fluctuations of up to 50% of the Green vote from one poll to the next.

remind remind's picture

Augustus wrote:
The Conservatives have the ability to start rising in the polls if they start focusing on the economy.

 

The main issue that concerns Canadians right now is the economy, not prorogation.  Most Canadians will be paying attention to how they are going to fare economically, rather than on the prorogation rallies.

And how do you know this if reality?

 

It would seem to be personal opinion, at best, given the turnout at rallies acoss Canada today. Moreover, Canadians can be concerned about several things at the same time, and democratic shortage being undertaken by harper et al, when campaigning for years on  democratic reform is high on people's list of awareness given the commentary aroud the net.

 

Quote:
Harper is already beginning to focus on the economic direction of the country:

 

Point?

As really,  it does not matter  what Harper has now decided to focus upon, Canadians want answers  in respect to why he and his government are dodging the Afghan detainee reality, as we are talking war crimes here and his possible making all Canadians be complicit in war crimes.

 

 

BTW your link did not work.

ottawaobserver

Sean, you may have missed the actual point of that post, because I was asking: if NOT prompting, then what ELSE.  You don't accept there's a gap, however, so that pretty much ends the convo between the two of us on that point.

Ken is right that I haven't heard another persuasive argument.  I'm not wedded to one thing or another, but if there is another methodological reason to explain why EKOS consistently has the Greens in the double-digits while no-one else does right now, I'm certainly all ears to anyone else who is not still bored with the conversation.

Finally, I asked what people made of the fact EKOS has started prompting with the "Other" option as well, and I notice no-one has answered that yet.  They register Other at 2.2% right now.  Independent and Other party candidates earned 1.1% in the last election, I hasten to add.

ottawaobserver

remind, Augustus signed up on January 4, 2010 and has done nothing but post pro-Conservative talking points since then.  I think we can take the hint on that one.

remind remind's picture

Oh okay, this is like the second post I read of his, though apparently Harper is not on board with "his  Canadians care about economic assumptions" either, as Harper stated today he was busy dealing with the life and death situations in Haiti.

 

 

ottawaobserver

Tonight, Ipsos-Reid (who usually underreport the NDP's numbers and overreport the Conservatives', according to ThreeHundredEight.com and the Paulitics poll index) have the following numbers out (change from Nov 19, 2009 in brackets):

N=1000, Tue 19 Jan - Thu 21 Jan

Cons - 34% (-3)
Lib - 31% (+7)
NDP - 17% (-2)
BQ - 9% (--)
Grn - 8% (-2)

The only regional numbers that are reported are:

Ontario:  Lib-38% Cons-37%
Quebec:  BQ-37% Lib-30% Cons-15%

Debater

ottawaobserver wrote:

Tonight, Ipsos-Reid (who usually underreport the NDP's numbers and overreport the Conservatives', according to ThreeHundredEight.com and the Paulitics poll index) have the following numbers out (change from Nov 19, 2009 in brackets):

N=1000, Tue 19 Jan - Thu 21 Jan

Cons - 34% (-3)
Lib - 31% (+7)
NDP - 17% (-2)
BQ - 9% (--)
Grn - 8% (-2)

The only regional numbers that are reported are:

Ontario:  Lib-38% Cons-37%
Quebec:  BQ-37% Lib-30% Cons-15%

I agree.  Ipsos-Reid does tend to have a pro-Conservative bias.  I have noticed it for 10 years now.  Back in the 2000 election in the week before the vote it was releasing numbers showing Chretien would only win a minority.  That turned out to be obviously false.

+7 is a big jump in Liberal support.  We'll have to see if they can maintain that momentum.  

The Quebec numbers look terrible for the Conservatives.  They are being beaten 2-1 by the Liberals in this poll.

ottawaobserver

For what it's worth, Darrell Bricker says that there's nothing much underlying the Liberal numbers except for a reaction to Harper.  Although I take most of what he says with a grain of salt.

Debater

It's probably true that the large majority of the change in the numbers is because of anger at Stephen Harper.

But that still doesn't change the basic fact - Stephen Harper has gone down in support and that's a nice thing to see.   Smile

NorthReport

The big perplexing question that most people continue to ask, if Harper is really so unpopular, is why are the Liberals not 20% to 30% ahead in the polls instead of being behind.

Bookish Agrarian

As a number of us have said the only important number is the Conservative one.  The trick is to knock down a sitting government- from there the votes park and it stands to reason they are parking in the Liberal lot.  I wouldn't get too excited if I were a Liberal though- the more voters see of Iggy the more the Liberals stand to lose parkers to other lots.

On the Conservative number this is good news and makes a fake budget and snap election less likely, not that I thought it was all that likely anyway.

NorthReport

I will be very surprised to see an election this year, but every Spring and every Fall we can count on the usual suspects in the media ramping up the issue.

Debater

NorthReport wrote:

The big perplexing question that most people continue to ask, if Harper is really so unpopular, is why are the Liberals not 20% to 30% ahead in the polls instead of being behind.

No major party in Canada is 20 or 30 points ahead of the other.  To reverse the question, if Ignatieff is so incompetent why are the Liberals not 20% to 30% behind in the polls?

Stockholm

ottawaobserver wrote:

For what it's worth, Darrell Bricker says that there's nothing much underlying the Liberal numbers except for a reaction to Harper.  Although I take most of what he says with a grain of salt.

 

I guess back in October when the Tories were at 40% he could have also said that there was nothing much underlying the Tory numbers except for a reaction to Ignatieff! Its easy to dismiss things and say that nothing matter and that nothing means anything - except that sometimes it does mean something.

Sean in Ottawa

ottawaobserver wrote:

Sean, you may have missed the actual point of that post, because I was asking: if NOT prompting, then what ELSE.  You don't accept there's a gap, however, so that pretty much ends the convo between the two of us on that point.

Ken is right that I haven't heard another persuasive argument.  I'm not wedded to one thing or another, but if there is another methodological reason to explain why EKOS consistently has the Greens in the double-digits while no-one else does right now, I'm certainly all ears to anyone else who is not still bored with the conversation.

Finally, I asked what people made of the fact EKOS has started prompting with the "Other" option as well, and I notice no-one has answered that yet.  They register Other at 2.2% right now.  Independent and Other party candidates earned 1.1% in the last election, I hasten to add.

OO I am perplexed by your first paragraph since I have been saying there is a gap for dozens of posts but that the gap is not coming from Greens saying they will vote Green and not doing it but instead from over sampling, poor methodology and self selection such that more Greens are counted than actually exist. This means there is one hell of a gap-- in fact in my opinion a much bigger one than most here are saying as I don't believe that there actually ever were 10% of Canadians contemplating voting Green-- it was polling methodologies and over selection and self selection that created that illusion.

I think this is a very important issue because if I am right then the strategy of trying to bring back Green support would be a horrible waste of resources as it may not exist in significant numbers beyond what we have seen in elections. Frankly, I respect your posts often but I am a little annoyed at this mischaracterization of my words because I have no explanation for it given how many times and ways I have said that I think the polls grossly over-estimate Green support and that it was never there to melt away by polling day in the first place. Disagree if you like but please don't misstate what I am saying.

ottawaobserver

OK, maybe I've identified the essence of our misunderstanding of one another, Sean (and, if so, hurray!). Is it that the "gap" you're discussing is the gap between what the polls in general are finding for the Greens, and what their "true" level of support is?

If so, then (and while I'm also very interested in that aspect) it's a little different than what I meant by "gap" in this particular case, which is the gap between what Nanos finds and what everyone else reports on the Greens' rated support, but most especially between Nanos and EKOS who always seem to have the highest numbers for the Greens (so high they don't seem realistic at all to me).

I agree with you that understanding the entire phenomenon has the utmost strategic importance. This is why I persist (and thus, no, it's not to aggravate you, per se ;-)).

I'm sincerely not trying to misstate what you're saying (it's certainly not my intent, and I really don't want to annoy you), but it is clear that we're having difficulty agreeing on what the issue definition is. In which case, no wonder we seem to be talking past one another!

Can you give this one more try, because I *do* understand the points you're making about an overall gap between Green headline numbers and actual support. Those points don't explain (or don't to me, perhaps because I'm missing a key fact, or assumption, or definition) the consistent gap BETWEEN the findings of 2 different pollsters. And I was looking for an explanation, if one is even possible, of what could explain that kind of gap. So far, when you only address the first kind of gap, I have understood that to mean that you believe the second kind of gap doesn't exist or is not meaningful.

wage zombie

I think Sean's explanation for the discrepancy between Green support in polls and actual Green votes (that Greens are being oversampled in polls, and therefore within those polls the support is exagerated) is quite plausible.  I feel like he's defined what he's talking about in elborate detail, and every few days he elaborates even more, finding new ways to describe his explanation.  I'm puzzled that such sharp babblers are still not clear about what he's saying.  Sean I feel for you in your frustration.

ottawaobserver wrote:

Sean in Ottawa says it's not the prompting, but EKOS has the Greens higher than anyone else.  Sean also says that Greens self-select into surveys more, but that doesn't explain the high (I think we all agree it's too high) Green number.

Yes.  It could easily explain the high Green number.  If Green supporters are more likely than other partisans to agree to be polled, then the Greens would be oversampled.  The sample for the poll would not be representative of the population, and so it would overestimate Green support--the poll would find Green support that's not actually there.

So ottawaobserver, you're wondering about the gap between Nanos and EKOS.  Nanos is so low, and EKOS is so high, and the only clear difference we can see is that Nanos does not prompt for party name.  The conventional wisdom is that people with no intention of voting will pick Green if the party name is prompted.  As a protest park.  Or something.  Yet, when not prompted with the party names, they will forget about Green and pick something else, maybe some other party or don't know or some other option.  This seems somewhat plausible.  I think Sean has argued that the actual way that polling happens (only a couple people called out of every 100 are willing to be polled) would make this scenario less plausible--because, for example, those agreeing to participate in a political poll would likely be politically engaged enough to remember the name of the Green party.

But you're more interested in explaining the gap.  EKOS is high, Nanos is low, Nanos doesn't prompt, therefore that explains the gap.  Nanos is always the lowest, it's the only one that doesn't prompt, therefore that explains the gap.  The problem is that you haven't demonstrated this gap.  The only number I see posted in here:

Sean in Ottawa wrote:

Nicky posted these numbers in the last thread:

Ekos 9.6

Angus Reid 7

Nanos 8.2

Harris 9

Strategic Counsel 11

Ipsos Reid 8

Segma 10

I think these are the #s from right before the last election.  So right off the bat we see there are two polling companies lower than Nanos--Ipsos Reid and Angus Reid.  Both of those companies prompt for the party name.  If prompting by party name adds so much to the Green support, then how come Ipsos Reid and Angus Reid are so low?  They should both be higher than Nanos.

I also notice that EKOS shows the 3rd highest support for the Greens, after Strategic Counsel and Segma--so your choice of EKOS seems odd.  There are 4 points separating Angus Reid (at 7) and Strategic Counsel (at 11).  Between Nanos at 8.2 and Ekos at 9.6, there's 1.4 points.

I know there was a recent Nanos poll out lately that showed the Greens really low, but i don't think you've demonstrated the gap that you're talking about.  If you want to say that Nanos is always the lowest, and they're the only ones that don't prompt, then you're going to have to show that, and you haven't.  You can't ask someone to explain a nebulous gap like that.

ottawaobserver wrote:

Also, do we have any idea why EKOS starting adding "Other" to the options they're prompting with?  Could it be they're trying to establish a baseline for "Other" in order to try a split sample (even if Sean doesn't think that's a problem)?

Possibly this might take some "none of the above" support in polls away from the Greens.  If Green support goes down in EKOS (and more of a decrease than is reflected in other polls) after this change then that would strengthen the idea that protest voters are parking support with the Greens.

ottawaobserver

wagezombie wrote:

I feel like he's defined what he's talking about in elborate detail, and every few days he elaborates even more, finding new ways to describe his explanation.  I'm puzzled that such sharp babblers are still not clear about what he's saying.  Sean I feel for you in your frustration.

Pardon me, buster, but Sean's not been the only one who's been frustrated.  I've never engaged in this conversation in anything less than absolute sincerity and good will, so really take your side-picking and cheerleading and triangulating and just tuck it away where the sun don't shine.  Please.

wage zombie

ok, sorry just was trying to help, guess i'll just leave you to it.

ETA: I never questioned your sincerity and good will.

thorin_bane

I understand seasn explanation but I just don't buy it(entirely). It goes some way to describe what we see just doesn't explain it though. No way we see the greens get 13% I will eat my shoe if they do. I also fail to see how the NDP have fallen to 14 percent as some polls have shown. There has been very little movement other than people getting pissed at the cons and deciding to not vote to push up the average percent of the other parties.

ottawaobserver

Well, I think I can see one thing I've done that's contributed somewhat to the confusion.  There's a missing word in the following passage.  I sort of assumed it, but I can see it might have been confusing to someone not inside my mind.

ottawaobserver wrote:

Sean in Ottawa says it's not the prompting, but EKOS has the Greens higher than anyone else.  Sean also says that Greens self-select into surveys more, but that doesn't explain the high (I think we all agree it's too high) EKOS Green number.

Now you make another comment, wagezombie, that I will try and concentrate on its meaning and ignore your condescending tone:

wagezombie wrote:

So ottawaobserver, you're wondering about the gap between Nanos and EKOS.  Nanos is so low, and EKOS is so high, and the only clear difference we can see is that Nanos does not prompt for party name.  The conventional wisdom is that people with no intention of voting will pick Green if the party name is prompted.  As a protest park.  Or something.  Yet, when not prompted with the party names, they will forget about Green and pick something else, maybe some other party or don't know or some other option.  This seems somewhat plausible.

There's another possible explanation, which is in fact much more plausible, namely that by prompting with the Green Party's name the pollster elevates their status in the eye of the respondent.  It is as if to say, if the Green Party were a viable alternative, would you support them.  Nanos asks a different way; he asks who the top two local parties' candidates you would consider supporting, or something like that.  Suddenly, you have to think of who's viable in your area, and see who comes to mind.  To me that sounds like a much more promising explanation of the difference between Nanos' findings and those of others.

It also seems to me that polls which prompt with both the party's name AND the leader's name, as some do (I can't say off the top of my head who does it and who doesn't, but I assumed others here might know, hence my original question), do show different patterns than polls which don't.

One thing I do know, and I believe Sean would concur, because I believe he's said as much previously (or maybe that was someone else), is that you can't compare polling firms' results JUST BEFORE AN ELECTION to those they produce now.  For one thing, the potential respondents are much more plugged in to the issues then because there's been a five-week election campaign going on.  For another, the firms know they are going to get rated on the accuracy of their last prediction before the election, and so have a special incentive to ensure their sample is especially representative and follows all the best practices, regardless of their usual methodology.

wagezombie wrote:

I know there was a recent Nanos poll out lately that showed the Greens really low, but i don't think you've demonstrated the gap that you're talking about.  If you want to say that Nanos is always the lowest, and they're the only ones that don't prompt, then you're going to have to show that, and you haven't.  You can't ask someone to explain a nebulous gap like that.

Listen, we all follow the numbers closely here (or I assume we do), and know where to find them.  If I haven't demonstrated the gap between Nanos and Ekos to you, GO AND LOOK IT UP.  You can go to PollingReport.ca (I believe I linked to them in one of my recent posts), or Wikipedia, or ThreeHundredEight.com, or the ElectionAlmanac.com, or the Paulitics Poll Index, or .... you get the picture.  You could go to the Nanos and Ekos websites for that matter.

If you don't accept that there's been a significant gap between EKOS and others over time, fine.  Both ThreeHundredEight.com and Paulitics have done studies on the various firm's gaps in favour and against each of the parties, that are available for consultation.  I do have some other claims on my time, and can't possibly do all that legwork for you, wagezombie.

I'm interested in learning from people who have worked in the opinion research industry what could account for such differences, hence my very sincere questions.  I'm really quite taken aback that someone would think it necessary to call me out for it.

ETA:  I see we cross-posted.  Without revising this entire post, I withdraw the parts that make me sound a bit snarky about your comments, wagezombie.

remind remind's picture

OO, I have to sidepick with you, ;) I just do not see what Sean is saying in this instance.

ottawaobserver

I don't want ANYONE to pick sides (although I always like to hear from you, remind).  Can't we all just have a nice socratic seminar discussion with one another?  The purpose of questions is to enhance understanding.  We need to find that zen again.

wage zombie

thorin_bane wrote:

I understand seasn explanation but I just don't buy it(entirely). It goes some way to describe what we see just doesn't explain it though. No way we see the greens get 13% I will eat my shoe if they do. I also fail to see how the NDP have fallen to 14 percent as some polls have shown. There has been very little movement other than people getting pissed at the cons and deciding to not vote to push up the average percent of the other parties.

The NDP is not at 14% and the Greens are not at 13%.  The polls are wrong.  That's the whole point.

It's not that 13% of Canadians will pick Green Party if the poll is structured some specific way.  It's that the sample is not representative.  That 13% Green Party support just does not mean much in terms of the voting preferences of the general population.

Pogo Pogo's picture

The self-selection theory as I understand it is that because polls only count those willing to be polled and that highly political people are extremely overrepresented in this group core support is over represented in the polling data. 

This theory does provide a justification for Green polls/votes discrepancies.  However, I am not sure it works in other instances.  The NDP polling vs voting data doesn't show a wide shift.  In particular back in the 1990's when the NDP support was at its lowest levels one would anticipate numbers similiar to the Greens but as memory has it the polls/votes were not widely different.

My pet theory is based on the difference between the general and the particular  People want to vote for the Greens, but on election day they cannot bring them to voting for a last placed candidate that they know is not going to win.  Conversely with the NDP back in the 1990's they overcame lacklustre support for the NDP in general to vote for particular candidates who were in tight races.

wage zombie

Here are the recent polls via pollingreport.ca

ETA: This didn't paste into babble like i thought it would and further attempts to add in colour for parties did not qork either.  SO the order is Cons, Libs, NDP, Bloc, Greens, and Other

Quote:

2010-01-21 Ipsos-Reid ...
34%
31%
17%
9%
8%
1%

2010-01-19 Ekos ...
31.5%
30.9%
14.9%
9.1%
11.5%
2.1%

2010-01-13 Angus-Reid ...
34%
28%
19%
9%
8%
1%

2010-01-12 Ekos ...
30.9%
29.3%
15.3%
10.2%
11.2%
2.3%

2010-01-05 Ekos ...
33.1%
27.8%
16%
9.8%
13.4%
--

2009-12-15 Ekos ...
35.9%
26.7%
17%
9.2%
11.2%
--

2009-12-13 Nanos Research ...
39.5%
30.2%
18.7%
7.7%
4%
--

2009-12-10 Angus-Reid ...
36%
29%
16%
11%
6%
1%

2009-12-08 Ekos ...
35.6%
26.5%
16.7%
9.9%
11.3%
--

While Nanos is the lowest, it looks like Ekos is more the outlier.

There are all kinds of different things that affect the poll--do these numbers demonstrate that prompting for party names adds significant support to the Green Party?  Maybe to some extent but it looks to me like Ekos is much more the outlier.

ottawaobserver wrote:

If you don't accept that there's been a significant gap between EKOS and others over time, fine.

I can accept that Ekos is high.  But Ekos being the outlier doesn't strengthen the argument that it's party prompting that makes the difference, unless Ekos prompts for party name differently.

ottawaobserver wrote:

I do have some other claims on my time, and can't possibly do all that legwork for you, wagezombie.

It's not my legwork.  You're asking Sean about a gap.  He explained the numbers he had.  Then he said:

Sean in Ottawa wrote:

Anyway I have answered the questions a couple times and there is no evidence and not even common sense supporting the theory that a 30-50% change can be explained simply by probing the name of a party.

So it sounds like he's not seeing the evidence you're seeing.

I already said that you are sharp.  I like your posts.  I'm just saying, you asked someone a question about a numbers gap.  You're likely to get a better answer if you clearly demonstrate the gap you're talking about, and so far you have yet to do so.  *shrug* This is not an attack.

ottawaobserver

Kindly note the last 4 Nanos poll numbers for the Green Party:

Firm | _Aug | _Oct_ | _Nov | _Dec
----------------------------------------
Nan | 4.6% | 4.6% | 5.9% | 4.0%

EKOS polls over that time frame (I'm eyeballing from their latest graph here, as I don't have all bloody night):

Firm | _Aug | _Oct_ | _Nov | _Dec
----------------------------------------
Ekos | 10% | 10% | 12% | 13%  (I tried to pick results from the same timeframe Nanos was polling in)

Shrug all you like, but there's a 5-10 point gap between the two sets of numbers.  Maybe if the word "gap" is causing the confusion between the two different uses of word in this conversation, then I will henceforth cause it a "discrepancy", but that seems to imply a value judgement, which I specifically did not want to make.

Bookish Agrarian

ottawaobserver wrote:

 We need to find that zen again.

 

 

Which is all I need for an excuse to post this-

 

 

A Zen master once said to me, "Do the opposite of whatever I tell you." So I didn't.

ottawaobserver

Perfect, BA!  Thank you very much for that.

ottawaobserver

Some more numbers to add to the discussion, this morning from CROP (N= ?, Que-only, Jan 14-24).  I ran across them in Norman Spector's online column for the Globe and La Presse, so they're not complete.

Best PM:

Ignatieff - 20% (a slide down from high of 45% in April)
Harper - 24%
Layton - 28%

Ballot question:

Bloc - 34% (-3); same as last September and down 4 points from Sept 2008 election
NDP - 17% (+5)

Sub-samples:

Francophones: Bloc-40%, Cons-20%, Lib-19%
Montreal: Bloc-32%, Lib-30%, Cons-15%
Quebec City: Cons-33%, Bloc-24%, Lib-22%
ROQ: Bloc-39%, Cons-25%, Lib-16%

The headline in the La Presse story is "L'effet Ignatieff s'évapore".  Even if you don't read french, you can figure out the meaning pretty easily.

I must say I'm getting pissed off with the quality of the media's reporting of poll numbers.  Basic facts are frequently not being included, for example the sample-size and the complete set of figures being cited.  And they don't seem to provide the complete CROP report anymore, which is quite annoying.

Pages

Topic locked