Friday, September 3, 2010

Funny things going on in SGI

Scott's DiaTribes offers up a riding poll from some pollster called "McAllister Opinion Research," apparently a Vancouver based firm, which surveyed 402 people and got these results Saanich--Gulf Islands, also known as Dear Leader's chosen seat:
Of the voters who have decided and say they are definitely voting, 34 per cent said they’ll support Lunn and 32 per cent will back May. Liberal candidate Renee Hetherington and NDP candidate Edith Loring-Kuhanga each received 17 per cent…
There you have it: 34, 32, and 17, 17. With a margin of error of 4.8%, that's a clear statistical tie between Gary Lunn and Ms. May.

But lets step back for a moment, and look at an analyst who got some pretty accurate results the other two times he attempted at guessing the end result in May's chosen ridings, and his take on SGI:

In case people are wondering, given all the data I’ve seen, and assuming a fall 2011 election (and assuming they continue their work at the current pace), this is what I’m currently projecting (note: a 3% range is about as narrow as I can get it):

CPC Gary Lunn 37-40%
LPC Rene Hetherington 25-28%
GPC Elizabeth May 21-24%
NDP Edith Loring-Kuhanga 9-12%

This comes from Greg Morrow of, whose projections in May's other ridings - Central Nova and London North Centre - were accurate for all candidates and their end result.

He also offers up an opinion on riding polls and their accuracy:
The problem is, as I’ve said time and time again, these high MOE riding polls are unreliable; they simply offer too little data, are rarely calibrated to the riding demographics, and never adjusted for turnout rates of demographic groups, and often don’t weight the data to the geographic distribution of population within the riding (which is particularly important in SGI, since May will run away with the Islands, but that’s only 13% of the population). The “too little data” claim is easy to see. If such a riding poll shows Lunn and May tied at 30%, what that really means is that May could be ahead by 10 points or Lunn ahead by 10 points; with a range of 20 points in the different possible scenarios, it’s just not that meaningful. They are, in short, a poor metric of projected votes. We’ve had riding polls days before an election that are so far off the actual votes, it’s not even funny (remember how J.P. Blackburn was supposed to lose Jonquiere-Alma by a landslide? um, not quite).
What I'm trying to say is, I don't believe the poll reflects reality. Morrow, in the same article, critiques the poor operation in SGI, the lack of identified voters and the push to do so, despite how much money has been poured into this riding (somewhere around $100K). Much like EKOS and their sometimes strange and unwieldy polls, something is off here. There is no way that May has that much personal popularity in a riding she's only lived in a few months, and apparently can't get her ground organization to run properly in.

My own personal prediction, going off the data that Morrow, Pundit's Guide, and Wikipedia can get me, is that Lunn ends up somewhere around 35%, May around 25%, Hetherington around 20%, and Edith Loring-Kuhanga also around 20%. Lunn simply has the benefit of entrenched support that won't be split along two other parties.

While on the subject of Elizabeth May, does anyone know why she didn't chose some more worthwhile ridings, like Guelph, Vancouver Centre or Bruce-Grey-Owen Sound? All of these have shown potential room for growth for the Greens, while SGI has consistently shown a 15% base for them. But what good is a base when it can't grow?

No comments:

Post a Comment