Finally, a new Poll, and more weirdness...

For the first time in three months, somebody decided to poll Albertans regarding their political views. Mainstreet conducted a fairly large survey of 2589 people across the province, asking them their opinions in both provincial and federal politics.

It is obvious by now that I'm an NDP supporter (by circumstance, rather than choice), and the initial results looked highly discouraging. According to their numbers, the NDP fell to third place among voters, with only 23% support. The Wildrose Party won the round with 38%, with the PCs finishing a strong second with 29%. The Liberals and Alberta Party, those two seemingly dubious entities, trailed significantly with only 5% support each.

Looking deeper, there were further surprises. The PCs lead the pack in Calgary, with 38% support, and the WRP is in a measely third - with 22%, outside the margin of error to be considered equal to the NDP's 26%. In Edmonton, the WRP are in a baffling second place, again outside the margin of error, leading the PCs 26 to 21.

There are some unsurprises. The NDP lead Edmonton with 43% support (with the closest competitor having only 26%, they're still in place for another sweep of that city in 2019). The Wildrose Party has a stranglehold on rural Alberta; with the PCs they would have 75% support (assuming naively nobody would jump ship in the process).

One issue with the polling data is that it relies on the outdated 2011 census for its weightings. The poll was conducted Feb. 9th-10th, days after the release of the new census data. It was undoubtedly too late to change the planning of the poll, but it is curious why they would plan their poll around so obviously important a date. Given that rural Alberta has continued its losing streak relative to Calgary and Edmonton over the past 6-7 years, we can safely assume the results over-represent the overall favourability of the WRP, while downplaying the governing NDP.

How would things really look if they used the 2016 data?

 According to the new census, 1.4 million people live in Calgary and area (which curiously excludes Foothills County and bedroom cities like Okotoks), 1.3 million in Edmonton and area, and 1.4 million in the rest of the province. One gets proportions of roughly 35% of the population living in Calgary, another 30% in Edmonton, and the remainder elsewhere. This is a part of a long term trend wherein Calgary and Edmonton gain more and more prominence compared to the rest of the province.

Given this, 43% NDP support in Edmonton turns into a base of 12.9% provincially. 26% in Calgary means one can add another 9.1 to that base - where we get the figure close to Mainstreet's result, 22%. But then, one needs to consider the remainder of the province. Weighed equally with Calgary, the 16% NDP support in the rest of Alberta would add a further 5.6% to the total. With rounding, one gets a result of 27.6% support for the NDP provincially - a fairly higher amount than one would be lead to believe.

For the WRP, their 48% lead in "rest of Alberta" gives them a base of 16.8%. However, they can only add  7.7% to that for Calgary; meanwhile Edmonton grants them a further 7.8%. Rather than having a 38-23 lead over the government, the WRP actually only has a 32.3%-27.6% lead; almost within the margin of error.

From the looks of things so far, I'm beginning to suspect the PCs are actually doing better than the polling suggested. Lets do the same test: Edmonton is a bust for them: 21% support gives a base provincially of only 6.3%. However, things do pick up! With the "rest of Alberta's" 27%, they go up another 9.45%; finally, with Calgary's significant 38% support, they get another 13.3%. In total, their provincial support can be estimated at 29.05%.

Now, I have to admit my process here is pretty sketchy. I simply don't know how Mainstreet defines these areas. You could consider this an educated guess, or wishful thinking, or data mining; whatever.

But I think we continue to have a horserace.

Mainstreet's Massive Sampling Bias

According to Mainstreet's own data, they surveyed 2598 Albertans for the results above (though the projections were based on a smaller sample of "decided" voters - 2,382 strong). From this respectable cadre of 2382 decided/leaning voters, however, we get this age breakdown:
18-34 year olds: 360
35-49 year olds: 590
50-64 year olds: 647
65+: 785

So, percentage-wise, these poll results were based on a sampling mix of 15% 18-34s, 25% 35-49s, 27% 50-64s, and 33% senior citizens.

This weighting simply does not at all reflect the population of Alberta; it doesn't even closely resemble the make up of Canada's oldest province, Newfoundland. Alberta is the youngest province in Canada, with a median age of 38, if I recall correctly. How is it that this survey so overwhelmingly samples the aged?

Maybe that old adage that polling is limited to people with landlines is truer than I thought (even though I've been polled on my cell, once).

Perhaps we'll only really know how people think when the votes are cast.

Thanks for reading.