Not all polls are equal


I’m learning that not all polls are equal, nor is the reporting on them free from editorial bias.

The three main methods for polling are IVR, Online and telephone. IVR is when a computer calls you and you enter your answer buy hitting a number on your phone.

Online polls (Ipsos, Angus Reid, Leger, Innovative Research, Abacus) rely on using a database of people that are signed up to do polls. So they tend to reflect the attitudes of people that would sign up for these things. This election year they have been largely consistent.

IVR pollsters (Forum, Ekos, Mainstreet) have been all over the map. One week the NDP are winning a majority. A few weeks later the Conservatives are cruising towards one. The response rate on an IVR poll is less than 1%. They have to call a lot of people to get anyone to answer.

Telephone polls (Nanos) are the most expensive to do but the response rate is much better at 9%. Nanos has consistently proven to be accurate when compared to the final results. They are generally regarded as the most accurate.

Some polls are properly weighted according to demographics (age, gender, education) and some aren’t. One IVR poll just released today polled 5000 people but only 400 between 18-35 and 2000 for 65+. That is going to skew results heavily towards Conservatives. While those 65 and over vote twice as much as the 18-35 group, the 18-35 group is 23% of the population and seniors account for only 16%.

If I were to get a sense how things turn out on election day I’d have to weigh seniors about 1/3 higher than the youth vote, not 5 times the youth vote.

Sometimes polls are released a number of days actually after they finished polling. The IVR poll I mentioned before finished on Oct 1st. Where three other pollsters have released much different numbers and were in the field in the last couple of days.

So who is winning?  I would guess the Liberals but not by much.  They need to be winning by at least a couple points to actually win more seats than the Conservatives.

Comments are closed.