Those who follow the polls may have noticed that sometimes we are around 9-10%, sometimes we are around 6-7%, and other times we are only around 5-6%.  What gives?  There are numbers enough for anyone who wants to trumpet that we are doing well, or that we are doing poorly.

I think the key thing to look for is the company doing the polling:  Decima Research has the highest numbers, Leger Marketing is in the middle and SES Research typically has us on the low end.  Each company will use a polling methodology that is fairly consistent from internal poll to internal poll, but is not necessarily consistent with polls from other companies.

What I suspect is happening is this:  in the 2006 election our high-end polling numbers solidified around 6-7% in the weeks prior to the election.  But on election day we only got 4.5% - about 2/3 of the high-end numbers.  This is the traditional percentage of votes that smaller parties like the Greens are able to convert to votes on election day.  A key reason for this is that we only have about 10,000 members, or an average of 30 or so members per riding.  So it is hard to implement effective get-out-the-vote (GOTV) campaigns in most ridings. 

Some of the polling companies are probably taking this into account, and some are not.  SES pollster Nik Nanos (who gives us the low-end numbers) spoke at the GPC convention in Ottawa in August and emphasised that GOTV is something that could give us big gains.  As our membership numbers and general level of acceptance by the Canadian electorate continue to increase, and we become increasingly able to develop and implement get-out-the-vote (GOTV) campaigns (as seen in the London North Centre by-election) the percentage of vote we are able to convert will increase.  It may well be that the low-end numbers from SES are already too low. 

For planning purposes, I suggest that the party take the number at the middle of the bell curve of numbers from all companies.  Or, since we like to focus on the bigger numbers, we should only use 2/3 of the high-end numbers from companies like Decima.  Then when we get to 15% from Decima or 10% from SES we can be more assured that we will hit double-digits on election day (and won't feel disappointed if we don't achieve the high-end numbers we are hoping for).

An aside:  When polling numbers are close news articles typically say the race are too close too call.  But I enjoyed this article from way back when from Kevin Drum that explains that a better way to think of poll numbers is in percentages, based on the margin of error (MOE) in the poll.  The chart from the article is below.  The percentages in the body of the chart represent the likelihood that candidate/party A is ahead of candidate/party B, given the percentage lead and the MOE: