03 June 2010

Williams party support drops nine points

Support for the Danny Williams Conservative party dropped nine points in three months according to the latest poll results from the provincial government’s pollster.

Corporate Research Associate’s quarterly poll showed that 58% of respondents indicated they would vote for the provincial Conservatives party if an election were held tomorrow.  That’s down from 67% in February.

cra may 10 The numbers are likely grossly inaccurate even with the correction presented here. The orange line shows the actual percentage of eligible voters who voted Progressive Conservative in the last provincial general election in October 2007. The blue line is CRA’s number, adjusted to remove their artificial inflation of Tory support.

The provincial government’s pollster doesn’t report the numbers this way, though.  CRA routinely inflates Tory support by as much as 28% by only reporting the percentage of decided voters.

These corrected figures also don’t account for the provincial government’s deliberate efforts to skew CRA’s polling numbers. As Bond Papers noted in late 2006, the Williams administration times its communications activities to correspond with their own pollster’s polling periods. probably one of the most significant examples of this would be the Premier’s disingenuous “have province’ announcement during the November sweeps month.

Local news media also routinely report CRA polls inaccurately by accepting at face value the CRA news releases.

Even allowing for problems with CRA’s polling, and for the government’s organized poll goosing efforts, that’s the largest quarterly drop CRA has reported for the Williams Tories since early 2005.



Cain said...

I'm not a statistician. I was having a hard time figuring out where your corrected numbers came from as compared to the numbers CRA reported. Eventually, I did come up with numbers that looked like your corrected numbers but the methodology I used to get them was so wonky that I must have stepped off the cliff somewhere. This is what I did.

In the May 2010 poll, 808 people in this province were polled. CRA reports that 75% of decided voters would vote PC but 22% of the total number of people polled were undecided/no answer/do not plan to vote/refused to state. So in order to know the total percentage, of the entire 808, who said they would vote PC, I subtracted 22% of 75 and then subtracted that number from 75% and got 58.5%. (75-22%=16.5; 75-16.5=58.5) Your corrected number was 58%.

I then did the same thing with the February 2010 CRA poll. CRA repors that 80% of decided voters would vote PC. 16% of the total number of people polled were undecided/no answer/do not plan to vote/refused to state. So, again, I subtracted 16% of 80 and then subtracted that number from 80% and got 67.2%. (80-16%=12.8; 80-12.8=67.2) Your corrected number was 67%.

Now this is where we jump in to the realm of "lies, damn lies and statistics". In the May 2010 poll, CRA says that 75% of decided voters would vote PC. That is as much as they could say given that 22% were undecided, gave no answer, do not plan to vote or refused to state. Your corrected number is also true to the extent that 58.5% of all the 808 people polled said that they would vote PC if an election were held on the day they were asked.

At least CRA does pretend to be psychic about the opinions of the 22% who really don't give a shit or were: eating their supper; watching the Young & the Restless; or trying to sort out the youngsters. (there would also be a percentage of the 22% who, like me, just hate pollsters, all pollsters). You, on the other hand, with your corrected number imply that none of the 22% would vote PC. That's just ... well ... "lies, damn lies and statistics".

But just for the hell of it, let's apply your corrections to the whole thing and see what we get.

PC = 58.5%, Lib = 12.48%, NDP = 6.24%, Undecided/no answer/do not plan to vote/refused to state = 22%.

Other interesting numbers from the May 2010 CRA poll are:

Mostly or completely satisfied with the Wiiliams Government = 87% (84.39% using your correction)

Who would you most prefer as Premier = 79% for Williams ( with your correction, 75.05%); 11% for Jones (with your correction, 10.45%); and poor Lorraine, well, she'd be better of changing her name to Undecided.

And lets not forget that all this polling was done in May. Which of course is well after the news of the shag up with the expropriation.

What is a poor humble e-scribbler to do???

Continue to emulate Fox with "fair and balanced" e-scribbling???

Nope. I say you should go buy some shares in Freshie. You'll make a killing in the next 10 years.

Ed Hollett said...

It's much simpler than that, Cain, although you gave it an able try at creative writing. Just go back and read the other posts and it is all clear.

Undecided is 22%. That means that 78% of the respondents were decided. If you multiply 78% by the number reported as the Tory percent you get the actual number of respondents who said PC (58.5%).

That 22% cannot be reallocated by CRA or anyone else. I don't reallocate them at all. As I noted in an earlier post, the method they use is shunned by pollsters because it distorts the results of the survey. The only way CRA could legitimately do anything with the UND other than leaving them alone is if they actually probed them and asked for any tendencies or leanings. They don't so CRA is doing a nose puller in its reports.

I take the UNDs at the face value of what they are, as they themselves allocated themselves.

Now to go a step further we know that CRA polling is actually off by about 50% when it comes to UND/Don't know/will not vote. In other words, CRA may get a figure of 15 to 20% when the actual figure on polling day will be upwards of 40%. (Look at the Straits, for example) That further undermines the credibility of their numbers and their methods.

You've then made another mistake in applying the correction to the satisfaction and Williams support numbers. If you looked at the tables of CRA figures, they don't actually skew those but give them as straight percentages. I didn't have a link to the current CRA poll but older ones are on the cra.ca website.

So since you failed miserably in your lame attempt to refute my point, I guess you'll have to try again. But here it is in case you got yourself buggered up:

The Tory numbers dropped dramatically this polling month, going from 67% down to 58.5%. That's the government's own pollster saying that, not me. That's a huge kick in the stones to any party and especially one that thrives on its polling.

What's more, they are basically back where they were before Christmas, at least which is on a steady downward trend since 2007. Not good, especially with the Old Man eyeing the exit every day.

But that's of course if you accept that CRA's numbers are:

a. accurate, (we already know they aren't) and
b. not skewed by other efforts to manipulate the numbers (we already know they are).

And it also doesn't tell you how the Old Man's party would perform against other potential leaders and candidates.

I am thinking it is a lot worse than the freshie-slurpers like to pretend. That pretty much explains the fevered references t polls these last few weeks among the bench-warmers in the House.

PoscStudent said...

Why haven't you given the NDP and Liberals numbers?

Ed Hollett said...

One simple reason:

Since all the media play will continue to be about torquing the interpretation of what are obviously dubious poll numbers, I wanted to focus attention solely on that.

The Telly published the CRA release word-for-frigging-word. bad enough they even ran the story given the obvioous problems with CRA's polling. To give them free space? Unconscionable.

CBC at least ran a graphic with the current corrected Tory number. (their math was off)

As I have maintained since about 2006, CRA routine inflates the Tory number artificially and for some reason flattens the Liberal number for some reason, even after you unbundle the whole decided/undecided thing.

As we saw in two by-elections last fall, the actual vote percentages were more like 60/40 or 55/35/10 or what have you.

There's some new research on the way, incidentally, which will likely have a significant impact on this whole business. We'll just have to wait until it is published.

Wm. Murphy said...

Since all the media play will continue to be about torquing the interpretation of what are obviously dubious poll numbers, I wanted to focus attention solely on that.

So you do not want to focus on the NDP and Liberal numbers because all the media are torquing the interpretation of the poll numbers!!!

Heaven forbid focusing or interpreting numbers on the other side...wouldn't want that!

Ed Hollett said...

Heaven forbid, Murph you might actually read something before randomly gainsaying. your point was covered in my answer to Cain's efforts.

Meanwhile, did you really mean to ask me if the government party can set government salaries?


PoscStudent said...

So how low are the Liberals and NDP really polling if the PC's support goes down so much?

Ed Hollett said...

If you look at actual voting results, you can see - as I have said repeatedly here and elsewhere - the provincial government pollster's results does some combination of the following:

a. artificially inflates the Tory voter choice result in the way the number is reported,

b. apparently boosts the Tory voter choice result anyway (maybe due to a huge problem with sampling or responses or as a result of government's deliberate campaign to manipulate the numbers (goosing))

c. flattens or reduces the Liberal choice, and

d. misses about 50% of the UND/will not vote category.

That would mean that Liberal numbers are actually higher than CRA reports, just as the UND/will not vote number is higher.

In Terra Nova there discrepancy was huge for the Lib and the "will not vote" number. Ditto the Straits which you could not have foreseen using CRA numbers.

The most obvious example of the problems was in the 2007 general election where the Tories garnered 43% of the eligible vote, not the 60 something CRA showed in its pre-vote survey or the equally absurd number in the immediate post-election survey.

All you have to do is compare what CRA says the numbers are to actual vote results to see the problems. Since CRA polls eligible voters you have to compare their results to that number, not to turnout.

Wm. Murphy said...

what about the numbers in Topsail?

Ed Hollett said...

Off as all the rest in one way or another.