Wednesday, March 25, 2009

Angus Reid Strategies Poll Results

The Headline result is:

  • Liberals - 43%
  • NDP - 37%
  • Greens - 13%
  • Conservatives 4%
  • Others - 3%

Angus Reid Strategies does a very different polling than other companies. They use people that have self selected to be to be Angus Reid forum panelists. They randomly select 800 people from this set and then adjust their results to reflect the population of BC. This is a new way to do things and is seen with some suspicion by many people in politics, I personally think it is valid if done correctly and I believe that Angus Reid Strategies seeks to be as accurate as possible.

The further breakdowns in the report are very hard for me to work with for a host of reasons.

  1. I do not know how many people out of the 800 stated they would not be voting, it should be about 300. This means I do not know what the margin of error is for decided voters in the poll.
  2. I have no idea how large the regional samples were, so I can not evaluate their relative accuracy.
  3. They do not provide data on how many people said they were absolutely certain to vote on May 12th. The numbers for the Greens look suspiciously high, they should have the largest drop.
  4. I do not know what the questions were that they asked
  5. I would like to see a lot more details on who responded and how they corrected the data. The more correction, the more the error in the results increase. If I could better understand how the data was developed I could do an analysis of it that I have comfort with.

The poll is clearly good news for the NDP, they are much closer than other polls have show and Campbell is clearly less popular than in the Mustel or Ipsos Reid polls.

If these results were to hold through to election day, it would indicate the Liberals winning 53 - 57 seats and the NDP 28 - 32. Though take those numbers with a grain of salt.

The difference between the Ipsos Reid poll and the Angus Reid Strategies does not mean anyone has a problem with their methodology. The headline results are ballpark within the same range, certainly not dramatically wrong.

So where does this leave us? I am only working with the headline results at the moment because of my issues with how ARS reports their results.

  • Liberals - 44.5% -1.3% from 2005
  • NDP - 36% -5.5% from 2005
  • Greens - 14% +4.8% from 2005
  • Others - 5.5% +2.0% from
I suspect that the Green and Other numbers will fall, but nearly as far as many people assume. I am also not at all convinced that there will be any strong move of people from the Greens to the NDP. The Greens polled in the same range last time as they are doing now, but they did so even in the polls were the NDP polled at close to what they got on election day.

Where do I think we will end up on election day? Now I am really predicting:

  • Liberals - 46.2%
  • NDP - 39.3%
  • Greens - 11.5%
  • Conservatives - 2.0% (based on 20 candidates)
  • BC Refed - 0.8% (based on 60 candidates)
  • Independents - 1.3%
  • The rest - 0.5%

So what does this mean for seats? More on that later.

6 comments:

Anonymous said...

One problem I have with the on-line Angus Reid Strategies polls is based upon the previous Alberta and Saskatchewan ARS poll results.

ARS tends to overstate NDP support and understate centre-right support.

Of the 4 pollsters in the last Alberta election, ARS came in last understating PC support by 10% and overstating NDP support by 4%.

In the last federal election poll of Saskatchewan, ARS understated CPC support by 14% and over stated NDP support by 9% with an ARS spread of 5% in favour of the CPC, when the actual result was a 28% spread in favour of the CPC.

Huge differences.

Today's BC ARS poll also has some apparent irregularities, which are inconceivable:

1. Greens - 13% "decided voters" and 13% "absolutely certain to vote";

The Greens have the softest base of all parties and that finding makes no sense at all.

2. Of the 3 income categories, the lowest income (< $50,000) show the **highest** level of Liberal support at 49% and the **lowest** level of NDP support at 33%;

That's impossible and goes against every other pollster's finding for that income category. In fact, the figures should be reversed.

Perhaps Evi Mustel said it best that "those that sign up for the online panels tend to work less and have lower levels of education than the national average."

Unlike the track records of Mustel and Ipsos, this poll represents ARS' third attempt at BC provincial politics.

I guees we will have to wait and see their poll results compared to the actual results on May 12.

Anonymous said...

Further to my previous post stating:

"2. Of the 3 income categories, the lowest income (< $50,000) show the **highest** level of Liberal support at 49% and the **lowest** level of NDP support at 33%;"

Those were the results when I viewed the online results this morning at 10:30 am when they were first released. Really!

But now those results in that same income bracket have changed since then to:

"NDP - 48%, Lib - 29%"

That's strange!

Anonymous said...

Anonymous 1:41 pm is unaware of Angus Reid's record covering elections with its online methods.

Best Pollster in the 2008 Federal Election (and closest prediction in BC) -

http://www.angusreidstrategies.com/uploads/pages/pdfs/2008.10.15_Election.pdf

Five Provincial Elections Predicted Within the Margin of Error -

http://angusreidstrategies.com/uploads/pages/pdfs/2008.03.28_Anniversary_1.pdf

And the Best Prediction in Quebec in 2008 -

http://www.angusreidstrategies.com/uploads/pages/pdfs/2008.12.09_QuebecElection.pdf

If that's not a track record, I don't know what is.

Anonymous said...

Yeah, the questions are on there.

Just look at the top of each resulting sheet.

Anonymous said...

Isn't there a problem with the ARS method that it favors "activists" and "extremists"? And this usually means union stewards and leftists/neo-progressives.

Imagine unions sending a memo to all their activists to sign up for ARS. They can easily take up 50,000 of the 100,000 slots and skew the poll. Even if they take up 20% of the slots - which I am sure this group has, that is the end to this methodology.

Glass

Anonymous said...

So where's your take on the ARS poll seat count?