Tuesday, October 09, 2007

Advice To Decima- Stop Publishing Quebec Poll Results

How can you take the latest Decima poll seriously, not the national number, that seems intuitive, but the regional Quebec numbers?:
But in Quebec, the Liberals were mired among the NDP and Green also-rans.

Some 35 per cent of Quebecers backed the Bloc, followed by the Conservatives at 26 per cent. The Liberals trailed with 14 per cent, the Greens were at 12 and the New Democrats at 11.

Now, compare those numbers with the one's released SIX days ago by Decima:
The overall numbers in that province show the Bloc Quebecois at 31 per cent, the Liberals at 23 per cent, the Tories at 22 per cent, and the NDP at 13 per cent.

I've already done a comparison of the two previous Decima polls for Quebec that show the same WILD fluctations, so the pattern seems clear. Can anyone (besides a boring Tory troll) explain how the Liberals number drops a full 9% in SIX days? It's not like the poll done SIX days ago was conducted during a period of Liberal bliss, as a matter of fact it was done in the midst of the "crisis".

Obviously, the margin of error is high for the Quebec numbers, that is the only way to explain the unreliable results. However, polls shape perception and I find it silly to tout results which show no relationship to results released SIX days ago. Unless an outfit devotes itself to polling enough people in a particular region to lower the margin of error, they should be prevented from releasing misleading results. We have seen the same ridiculousness with Decima, as it relates to British Columbia, where parties see upwards of 16% swings between poll, hardly. In other words, I don't see the value in everyone digesting something which has proven unreliable. Poll more people, or don't release partial regional numbers, because they actually influence decisions.

As it relates to the national numbers, I don't find the gap surprising, given the barrage of negative coverage. As a matter of fact, the "still tied" poll was more of a head scratcher. I don't see this result as much of anything for Conservative partisans. Still below the election result, despite the Liberals at rock bottom, not exactly champagne time.

5 comments:

A View From The Left said...

I have to admit that I always take Decima's Quebec regional with a grain of salt as more often then not it seems to be out of wack with all the other polls. How much the wonky Quebec numbers affect the overall polling numbers is debatable.

Personally, I'm waiting for the next SES poll which I think should be due out sometime soon.

MarkCh said...

"Should be prevented?" Watch out, your mask slipped. Of course the margin of error is large - I am no statistician, but I would be surprised if it were much less than 10% on a regional breakdown. Still, that's free speech.

Steve V said...

"Still, that's free speech."

Of course, but when the media runs with the numbers and concludes "Liberal free fall", it creates a dubious distinction, based on shaky results. Polls have influence, and they should be held to a certain standard.

MarkCh said...

held by whom?

Steve V said...

The media is supposed to have journalistic standards, and they decide daily what is news worthy and what isn't. If you have a poll that completely contradicts the results from SIX days ago, then there is some responsibility to acknowledge the inconsistency in the interest of fair reporting. You are trying to attach this "big brother" angle which is frankly nonsense, standards are a common occurence, in every discipline, that has little to do with quelling "free speech".