Latest polling thread - started April 10, 2012

Excellent - that's the poll that was discussed tonight on P&P.

I'd take every conclusion 308 comes to with a huge grain of salt, particularly at the seat level. Their record in predicting the last federal election speaks for itself. Much like the pro-strategic voting sites, 308 tries to make numbers say more than they really say. Also, I contend that the last Liberal votes that the NDP will shake loose from the Liberals are in Liberal held seats, so I wouldn't lose too much sleep over NDP gains. The main issue I have about 308 is their slavish adherance to a flawed model over common sense.

The monthly averages I do like though, and as the last sentence says: "A last snapshot, then, of the political landscape in Canada before the arrival of Thomas Mulcair." All polls since Mulcair took the leadership show the NDP and Cons statistically even, so there would have to be some pretty wild vote distributions for this to result in anything other than a minority parliament.

Another pet peeve I have with 308 is that they completely don't look at momentum, in fact they provide so much "hysteresis" that they are actually doing the opposite of considering momentum. This does avoid showing false change, but like in the case of the last election it also made it look like the shift to the NDP wasn't really happening until much after it started. Astute political observers understood that when the NDP passed the Liberals they would most likely keep going up by peeling off strategic voters. (regardless of how strategic their vote really was), whereas 308's models were still looking at what happened 2-3 weeks earlier.

So that said, I predict that most of the new polls will show the NDP leading in the near future, and I think we've only seen the beginning of the Mulcair bounce. If I was a Liberal I would be desperately trying to figure out a way of reversing this trend.

Yeah,

here I made line graph on MS excel

May 2, 2011 to April 4, 2012

Better than Wikipedia fed poll firm graph?

Cons going to dive, NDP going to tsunami wave, Libs going to die like flat, and finally Green/BQ are usually neck and neck but Green beats BQ

Lol, its Dark Orange looks like Red colour.

That what i use MS excel "Gradient line colour" not Solid line.

Thanks for the graph, Ippurigakko, love it!

Not always  but especially during elections they are worth taking with a grain of salt

It is very easy to frame the questions to get the answer you want

Many foks like to vote for winners for some reason...so if they see poll projecting certain party to win it influences their vote...

Sorry, I don`t agree with you, NorthReport.

If you really think a reputable market research company changes the real numbers, then you must never believe the results of any polling. If this is the case, this polling thread is meaningless, isn`t it?

Of course, different market research companies may show different results for the same kind of polling and this is normal. There is no 100% correct polling. 

As I read NorthReport's comments, he spoke of believing polls more when close to elections.  I find polls during elections, when the election date is more than 10 days away, to be suspect.  Polling firms that do a lot of government business have too much vested interest for me to trust them. Having analyzed night by night results of NANOS during the last election, for example, shows that the results of some of the nights were so improbable as to be statistically impossible, not once but several times.  It is telling that similar analysis close to election night shows no "beyond the realm" of possibility evenings.  NorthReport is correct when he says that there is no way of "proving" that the polls are manipulated unless it is very close to an election date.

In between elections, I am more likely to trust those firms without strong connections to one political party or another.  There have been too many cases of "push polls" undertaken by "reputable" firms for me to trust results all the time.  Since the Liberal Party has fallen to third place, I tend to trust EKOS more now as there is no obvious connection between company government business and the polling results.  I trust the Conservative pollsters less now as there is such an obvioius connection.  I distrust NANOS polls the most.  I tend to like Harris Decima in spite of the company's Conservative connections.

On the other hand, if an election is due within a few days, I trust that all reputable pollsters are trying to get their polling result to reflect what the actual outcome will be.  Therefore, there is a time when I have a lot of trust in pollsters.

"Do you not agree that is is very possible to skew your poll results based on how you pose the questions?"

Of course, it's very possible to skew poll results. That`s why I said "reputable" market research companies.

Polling can be skewed in many different ways. 

A simple example is for a pollster who wants to show a particualr political party looking good. All the pollster has to do is overpoll in a geographical area that is favourable to that political party. Or if they want to show a political party looking bad, all they have to do is do the opposite, and underpoll in areas that are favourable to therm.

But regardless of whether or not we can believe polling results, they can have a huge impact on the voting public, and that is why there are so many of them released, particularily around election time.

There has to be a bit of a spread in the % before the NDP pulls ahead in the seat count.

Headline on www.cbc.ca/politics webpage, Liberal MPs say F-35 buy bungled. No wonder Rae is rated highly. I am sorry, but I really hate the CBC, and the rest of g-d MSM. When are they going to figure out the Libs are the OO and Rae isn't leader of the OO? I mean, really!

Seat counts over-represent parties going down and under-represent parties going up as a general rule. And I mean this in terms of long-term trends.

I think it is partly because the over-value incumbency but there is another reason. When a party is very low their support is more evenly spread but as it increases peaks and valleys appear. A party that is low in an area and moving up will according to the model keep the same flat appearance and not have that gain reflected in seats but when a party in reality comes up and the peaks appear those will become gains in seats the model did not see. Problem is you cannot predict where they will be-- there is no possible model for that as the historical baseline is flat.

By the same token parties going down are expected to keep the same peaks and valleys when in reality they tend to flatten out. There are exceptions of course with candidates who are elected not because of the party but because of their personal popularity.