Thursday, June 15, 2006

Polling revisited

Well, the proof of the point we were making a few weeks ago about the unreliability of public domain polling has been made clearly for all to see next door in Ontario.

Manitobans may have seen the recent headline about Dalton McGuinty's Liberal government being 8 points ahead of the John Tory, well, Tories in an SES poll conducted in late May.

What you may not have seen was a poll taken at almost exactly the same time by Leger that puts McGuinty's Libs 2 points behind Tory's Tories.

This disparity is described here.

Both companies are established market research firms. Both had significant sample sizes (about 500 in the SES poll, about 1000 in the Leger poll) as well as following established sampling and opinion research methods.

Both informed us the margin of error was minimal (4.5% -- SES, 3.1% -- Leger). The spread between the two polls is so wide it cannot be explained away by these margins.

The danger with the Probe/Free Press situation here in Manitoba is they are the only game in town. And their results are picked up and repeated by their polling-starved competitors. No one else is doing public domain work to show that different polling firms often do things quite differently and come up with significantly different results.

There is a lot of art to polling. Unfortunately, journalists tend to overly rely on them to generate copy and they do not convey the alchemy aspect to their audience very much.

Some have said this is because they don't understand polling particularly well. We, however, have no particular opinion on that issue.