Democracy Watch thinks the media are doing Canadians a disfavour by not giving the undecided vote when they report polling results.

This was sent out as a group email on Friday:

Dear Editors/Producers,

As the federal election speculation rollercoaster continues, many media reports are misleading the public concerning public support of the federal parties, and hurting the media's credibility at the same time.

The reporting of the results of about a dozen polls in the past month has often neglected to mention one key fact -- the percentage of voters who told pollsters they were undecided. Any media report that leaves the impression that pollsters found no undecided voters is misleading.

Assuming that the percentage of undecided voters is about 20-25 percent right now (as it usually is at this time before an election), and given that the gap in the polls between the Liberals and Conservatives has been no greater that five percent (when polling error rates are taken into account), the recent polling results could all be summarized as actually showing nothing important in terms of possible election results.

If my assumption is wrong, please let me and all Canadians know. Hopefully sooner than later all Canadian media outlets will present a summary of recent polls that includes the percentage of undecided voters, and seriously takes into account the error rate, so that Canadians will be told the actual reality of the pre-election situation.

Or are we headed toward a repeat of the polling reporting fiasco of the June 2004 federal election?

Sincerely,

Duff Conacher, Coordinator
Democracy Watch
P.O. Box 821, Stn. B
Ottawa, Canada
K1P 5P9
Tel: (613) 241-5179
Fax: (613) 241-4758
Email: dwatch@web.net
Internet: http://www.dwatch.ca

My thoughts

The polling fiasco of which Mr. Conacher speaks is the difference between the final poll of the 2004 election campaign and actual voting-day results (here's a CTV.ca story).

Ipsos-Reid, CTV's pollster at the time, had the breakdown as follows on June 25 (the actual vote on June 28 is in brackets):

Liberals - 32 (36.7)

Conservatives - 31 (29.6)

NDP -17 (15.7)

Greens - 6 (4.3)

1.4 + 1.3 + 1.7 = 4.4

The Liberal vote went up 4.7 percentage points. Given the dynamics of that campaign and the fear the Liberals were able to create of a Tory victory, a major voter migration to the Liberals from the NDP, Greens and very soft Conservatives explains the poll result to me.

And remember, while the poll was published on the 25th, data collection ended on June 23. That gave voters four days to think it over before they went into the booths.

With our electoral laws, no polling information can be published less than 72 hours before a vote. The pollsters claim they could see the vote shifting, but were unable to report anything.

Obviously, as the number of undecideds goes up, a poll's potential reliability declines. Voters usually harden in their choices by election day.

Different pollsters use different assumptions to account for the undecideds. Some will assume a certain percentage simply won't vote. Others will assume the sample is weighted against the incumbent. And yet others will ask questions that try to determine which way the respondent might be leaning.

One could also assume that once the undecideds choose whom to support, they will vote pretty much in the same proportions as decided ones do.

In terms of current polls published by the news media, they all seem to be relatively close, with a slight Liberal lead but the Conservatives close behind.

If there were as much volatility as Conacher suggested, the polls should be all over the place. But generally speaking, we aren't seeing unexplained -- or unexplainable -- volatility (the Liberals dropped in the immediate aftermath of the Gomery report's release, for example).

While the level of undecideds would be a useful piece of information, I don't believe the news media has been misleading people by not reporting it.