Crawl Across the Ocean

Wednesday, May 20, 2009

Credibility Watch

One of the good/bad things about blogs is that it's all in writing, so you can often go back and see what was written in the past by various people. Occasionally I like to take advantage of this and cherry pick some recent event, where I was right in some prediction and others were wrong, in order to highlight my superior predictive abilities.

For instance, here’s a recent post by Andrew Coyne regarding the polls leading up to the B.C. election – the post was written before the B.C. election took place, so now that the election has happened, we can assess whose interpretation of the polls was more accurate:

Andrew wrote:

I dunno. This makes no sense to me:

"Halfway through the B.C. election campaign the race for the leadership is tighter than ever. The BC Liberals and New Democrat parties are separated by just three percentage points, according to a new Angus Reid Strategies/CTV poll.
The online survey of 822 voters found the BC Liberals are holding top spot with 42 per cent, with the opposing NDP not far behind at 39 per cent. The Green Party remains in third place with 13 per cent."

The same poll shows that, by a wide margin, the public thinks the economy is the number one issue. And that, by a margin of three to one – 48% to 16 — they pick the Liberals’ Gordon Campbell as the best person to handle it. Nearly two to one — 40% to 23% — also prefer Campbell as “best premier.” Yet the Liberals are only three points ahead?

Going into the campaign, the Liberals enjoyed a 17-point lead. Nothing much has happened since. The NDP have run a scattershot, unfocused campaign. And suddenly they’re only three points behind?
Colour me skeptical.

And then in the comments:

dan in van says:
Smacks of a push-pull that was manipulated, whether accidentally or purposely, to ensure that Liberal forces are motivated to vote. It is the only poll in a few years that have shown this kind of narrow margin. But while my theory may sound a little conspiracy-like, in any event I take this with a long tall one.

Critical Reasoning says:

I never trust online polls like this one. Here’s the justification by CTV/Angus Reid:
Our polls are conducted online, and not over the telephone or through face-to-face interviews. The rationale behind this approach is simple. Canada is now one of the most connected countries in the world, and Internet access has spread to every key demographic group in the country, including seniors and people with below average household incomes.

Conversely, telephone-based research has been greatly affected by declining response rates and the absence of land-lines in the homes of many younger Canadians.
Codswallop! Online polls are only used because they’re cheap. Almost any online poll is more likely to magnify the young, left-leaning “can’t be bothered to vote” demographic at the expense of the older, more conservative demographic who always vote, but only venture online to check email.

Derek Pearce says:
Heh. Where can I online-vote that Unicorns are the prettiest of all the fabled creatures? Actually I bet that you’d get a better than 19 out of 20 result for that, so it’s a bad example. What was the methodology of survey even if online? If people had the option of clicking one party or another from a public site and meandering off, then the NDP will be over-represented by zealots who did their quick-click duty and then went web-surfing elsewhere, while those who cared stayed to answer more in depth questions will more accurately reflect the public opinion of Campbell.

bigcitylib says:
A response to the speeding scandal, which arrived plop in the middle of the polling period?

As for AR’s online polls, I’ve complained alot about them but during the last Federal they did as well as (maybe better than) anyone else’s.

catherine says:
These results are only a small shift from the last Angus Reid poll (done in March which put the Liberals at 43 and NDP at 37) and the small shifts could be fluctuations in small numbers (MOE) or due to the speeding scandals.

Mike G says:
Well, no one ever accused British Columbians of being sensible.
Here’s the thing, though: I would contribute to those statistics. I think Gordon Campbell is probably a better administrator than Carole James. I am not entirely happy with either platform; the economy is certainly the biggest challenge ahead, but I don’t actually think that the provincial government can do much about it. I’m very unhappy with a lot of the things that the Liberal government here has done over its current tenure, I think there’s been a lot of hamhandedness and ignorance on their part, and I think a bit of shakeup could be good for them. On top of that, my local MLA, Shane SImpson (NDP, Vancouver-Hastings) seems to be just fine and I see no need to boot him.

Dale says:
The 17 point lead poll is a Mustel poll. This company was started by a current MLA for the Liberals. If any poll has a tendency to bias it is that one.
The A. Reid poll is very close to Robbins Reasearch poll wich uses telephone polling. You can complain about one but it is backed up by the other. The BC election will be won or lost in each riding. Liberals are strong is some but the NDP are equally strong in others. Who ever gets the vote out will win.

Declan says:

Libs won by 4% last election. Seems pretty plausible to me that things haven’t changed much since then. I guess we’ll see soon enough.


Election result:
Liberals get 46.02% of the vote, NDP gets 42.06%, Liberals have a margin of victory of almost exactly 4%.

Credibility Increase:
Angus Reid, Me, Catherine, bigcitylib, Dale and Mike G.

Credibility Decrease:
Mustel, Andrew Coyne, dan in Van, critical reasoning and Derek Pearce

That is all. Just trying to enforce a little accountability around here...

Labels: , , ,


  • What do you think of Robbins Research? They generally seem to have faulty numbers (especially the MOE for given confidence levels), and I have a suspicion that they're outright fabrications. (see this for example - that MOE is too low for the sample size)

    By Blogger JG, at 7:06 PM  

  • Touché. It just seemed weird to me.

    By Blogger AC, at 8:26 PM  

  • Josh - Aside from my contempt for Ipsos Reid (see here I don't really have an opinion on the pollsters, although Glen Robbins, CEO of Robbins Research, stopped by to semi-spam my blog back in the day which was a bit odd.

    AC - well, to be fair, it's easier to have an idea what makes sense while living out here. At any rate, I'm sure your pleasure at Mulroney's shenanigans finally starting to catch up with him outweighs any dismay at misreading a BC election poll :)

    By Blogger Declan, at 10:37 PM  

Post a Comment

<< Home