An abundance of polls were released into the media trough today. Political observers are pigging out. It’s hard to make sense of the numbers amidst the contradictory squeals – so here’s a quick breakdown below.
Four polls, four plots. Different methodologies, sample sizes, and outcomes.
Mainstreet was the best for CPC with an 8-point lead. Conducted September 30-October 1 it is arguably a bit stale but does have a very large sample.
- Standout stats – Has CPC with a 7-point lead among women. This is very different from Nanos which has a 13-point lead among women for Liberals. Has 10-point CPC lead in Ontario while others have marginal lead for Liberals.
Ekos shows a 4-point CPC lead. In field October 3-5 with combined IVR / live phone methodology (about 2/3 IVR) with second largest sample size.
- Standout stats – Conservatives tied for lead in Quebec with NDP at 28%. NDP with 9-point lead in BC. 60% for Libs in Atlantic is highest by 10-points.
IPSOS shows a dead heat. In field October 2-5 with its online panel.
- Standout stats – NDP closest to the front of the pack than any other poll – at 7-points. IPSOS has an almost even gender split for CPC and Libs, which is uncommon. Libs tend to do better with women, CPC with men. Lib 17-pt lead among under 35’s would suggest turnout challenge, if accurate.
Finally, Nanos has the Liberals up by almost five points over CPC. A gathering trend over the past few days. Nanos conducts 400 live telephone interviews each day, combining them into a rolling track of a sample size of 1200.
- Standout stats – the national race is THE standout stat. In addition, the CPC appear low in BC at 24% compared to 35% for Libs. Libs have an 8-point lead among 50-59 year olds, sandwiched between CPC leads among 60+ and 40-49s.
What’s the upshot?
Only one conclusion – the NDP are huffing and puffing but can’t seem to blow down the two-way race (trying to work the pigs back into the story).
The sub-samples are inconsistent between the pollsters, which is commonplace considering the margins-of-error increase when you raise the hood. There are many other sources of potential unreliability that relate to ongoing problems with pollster accuracy.
Here’s two previous posts that should give observers pause for thought: