Tuesday, November 10, 2009

Last Word on Faulty Poll

As there was some legitimate criticism of my post about the poll published in La Presse on the Friday before the Montreal municipal election, I'd like to make some final comments.
First of all, when I wrote my post the page linked was not available. I'm not that stupid.

After reading the comment by Fagstein and following the link, I am still of the same mind that the poll was faulty.

I respect Fagstein as one of Montreal's finest bloggers (specializing in the domain of media), however I do believe I have a better insight when it comes to polling.

For twenty years as an independent political organizer, I conducted countless internal polls and I recognize when a poll is blown.

The poll in question was published on Friday, from information compiled the previous two days. The data was current, you can't do better.

The margin of error is provided to account for all the excuses that Fagstein uses to justify the erroneous results. If we'd accept that polls can be off that much because of what he describes, then no poll could be ever considered reliable. However most are. Professional pollsters know their stuff.
If we are to believe the poll, Bergeron dropped 25% and Tremblay gained 25% support in three days.
Not unheard of, but extremely unlikely.

Pollsters admit that for various reasons, about one in twenty polls are wrong to the point of badly distorting reality.
This may happen because of faulty methodology, interpretation or a sudden whirlwind shift in the opinion of the sample subject. Sometimes it is plain bad luck, where pollsters hit a large block of respondants who are unrepresentative of the general group.

Sometimes events change between the time the poll is taken and when it is scheduled to be published. In that case professional pollsters trash the poll entirely.

The fact remains. The poll was terribly wrong. In the case of the support for Mr. Bergeron and Mr Tremblay, the error was double the margin that pollsters provide themselves.

That in itself is newsworthy. La Presse should have have addressed the issue, especially since they were accused in the media of being anti-Tremblay.
Post-election, Simon Durivage, of the RDI news network commented on air "It's clear that La Presse lost it's election" ('c'est clair que La Presse a perdu ses élections").
Yves Boisvert of La Presse bristled at the barb and took Durivage to task in a column. Perhaps he should next comment on the blown poll.

The main point of of my post was that my prediction that Mr. Bergeron would win the election was based on a faulty poll. I stand by that statement. As for a conspiracy, I stand corrected.

Do you think polling is always honest?
Here's a classic and humorous example of a 'push' poll, a poll designed to elicit a certain response and meant no to gauge public opinion, but rather to shape it.

No comments:

Post a Comment