Does May have a chance in Saanich — Gulf Islands?

Green PartyShort answer:  Absolutely, yes.  She has a very good chance.

Long answer:  Still probably yes.  However, if she runs her campaign the way she ran her candidacy “announcement”, then she could be in serious trouble.

Elizabeth May seems to have definitively chosen British Columbia’s Saanich—Gulf Islands riding as her choice in which to run in the next federal election.  I have to admit that I am intrigued by May’s decision to run there.  The circumstances around this move have been less than ideal for May and the Greens, however I don’t think it’s a bad move (though the data clearly shows it probably wasn’t the best move if you take “best” to mean “highest chances of success”).  This move possibly could very well end up paying off for the Greens (more on that later), however the manner in which May has done this has been amateur hour at the comedy club.

First the downside for May and the Greens, then I’ll end on an upbeat note.

In speaking to media earlier this month, May stated that there was still a variety of issues not yet resolved within the party regarding her candidacy in Saanich—Gulf Islands.  However, in the same breath, she said that Saanich—Gulf Islands was her choice for where to run.

I guarantee you May’s political operatives cringed when she gave that quotation.  This raises several questions all of which reflect poorly on May:

  • If she comes out and states that this riding is her choice, then why is the Green Party senior organization still “considering” it?
  • Has a rift broken out between Elizabeth May and the Green Party senior organization?
  • Why is Elizabeth May announcing this while a process to decide which riding to select is still ongoing?  Wouldn’t this announcement render such a process moot?

Furthermore, the Greens already have an announced candidate in the riding who hails from the left flank of the party and does not appear willing to step down.  May’s party is one of the most ideologically diverse in the country, and in such situations, it is generally important to keep all wings of the party happy.  May’s hasty move now risks stoking the ire of the left wing of the Green Party which, when combined with the criticism May’s been receiving from the right flank of her party, makes finding that winning dynamic that more difficult (though not impossible).

May needs to learn that there’s a reason why serious politicians don’t give a solid answer to a question when the decision is not yet firm and all of the ramifications have not been fully itemized.  This is a pretty remedial lesson in political circles and it usually comes around the same time that political operatives learn that decisions don’t get made in committees and that you never ask a question in front of a microphone to which you do not already know the answer.

Moving on to the good news for May and the Greens:

Despite all of this, the data still show that May has at least a decent chance of snatching this riding from the clutches of the Tories.

Greg Morrow over at Democratic Space has a post analysing whether Elizabeth May can beat Gary Lunn in Saanich—Gulf Islands.  Greg concludes that May doesn’t have a chance against Lunn and that even the best possible outcomes have her loosing by several thousand votes.

I think Greg is a fantastic political analyst and I have a great deal of respect for him.  However I’m going to have to respectfully disagree with his conclusion in this matter.

So far May has run in two elections (London North Centre in a 2006 by-election and Central Nova in 2008).

In the 2006 general election in London North Centre, the election results were as follows:

London North Centre 2006

When May ran in the by-election in London North Centre, the results were as follows:

London North Centre 2006 Elizabeth May by-election

As you can see, May improved the Greens’ showing by 20.3%.

As for Central Nova, in the 2006 general election, the results were:

2006 Central Nova

When May ran in 2008, the results were:

2008 Central Nova with Elizabeth May

As you can see, May improved the Greens’ showing this time by 30.5%.

From this data, we can extrapolate a few things.  When Elizabeth May runs in a riding, on average the following happens:

  • The Greens go up by 25.4%
  • The Conservatives go down by 1.3%
  • The Liberals go down by 5.3%
  • The New Democrats go down by 12.8%

However, the overall Green vote increased in Atlantic Canada by just shy of 4% between 2006 and 2008.  So roughly 4% of Elizabeth May’s increase wasn’t due to her presence but rather was due to an overall rise in Green Party fortunes in that region.  Because of this, we need to reduce her number (25.4%) by roughly 4% to fully take into account just the amount that Elizabeth May’s presence increases the Green vote in a riding isolated from other factors.  Doing this, we end up with the following data:

  • The Greens go up by 21.4%
  • The Conservatives go down by 1.3%
  • The Liberals go down by 5.3%
  • The New Democrats go down by 12.8%

Now, we need to apply these numbers to last election’s results in Saanich—Gulf Islands.

But before we do, there is one problem.  Last election in Saanich—Gulf Islands, the NDP candidate had to drop out after his skinny dipping incident came to light.  We know that whenever an NDP candidate drops out, it has an effect not just on the NDP’s total numbers but also on the other progressive parties’ numbers as the disaffected New Democrats “park” their vote elsewhere.  Thus, assuming that next time the NDP candidate in this riding is unlikely to become disgraced again, I will be using the 2008 election results for the Conservatives and the 2006 numbers for the Liberals, NDP and Greens.  (Note that, if anything, this will provide a slight bias in favour of the Conservatives as their vote total was slightly higher in 2008 than in 2006).

So we have this as our base:

Saanich Gulf Islands base 2006-2008 numbers

Applying the “Elizabeth May change numbers” to this base projection, we get:

Saanich Gulf Islands Elizabeth May Projection #1

Now that’s still a win for the Conservatives (and a safe one at that).


Remember that we earlier stripped the change in the Green Party’s vote between 2006 and 2008 out of Elizabeth’s May’s numbers.  So now we have to reintroduce the change in the Green Party’s (and the other parties’) polling numbers but this time we have to do so looking at the region of British Columbia rather than Atlantic Canada.

It turns out that the Greens have an overall more favourable climate in BC than they do in Atlantic Canada and this is important.

Beginning from the “raw vote projection” just calculated above, we now will apply the changes to each party in order to obtain our final projections.

Saanich Gulf Islands Elizabeth May Projection #2

Now, I do not vouch for the exact vote totals appearing in these final two columns.  However, as a rough predictor of how probable an Elizabeth May victory is, I think this is a strong indicator that it is entirely possible.


Greg Morrow below noted that I forgot to take into account the fact of the change in each party’s support between the polling data and the actual vote results.  I thank Greg for pointing this out because this is an important revelation considering that the Conservatives tend to do better in elections than in polls whereas the Greens tend to do much worse.

vote change

So, applying these changes to the above above projections, we get the following “final final” projections (as best as we can given the sparse data available):

New Arithmetic

New Geometric

So we have a much tighter race in the Arithmetic projections here.  Furthermore, the race could definitely slide back into Gary Lunn’s hands if the Conservatives can retain their 2008 vote/poll discrepancy rather than what they witnessed in 2006.  The race could also slide back into Gary Lunn’s hands if the Greens get closer to their 2008 vote/poll discrepancy rather than what they witnessed in 2006.  However, my theory is that the vote/poll discrepancy for the progressive parties witnessed in 2008 was partly exaccerbated by the fact that the vote was held so early in the academic school year which meant that all young 1st year university students studying in a different town might not have yet received the two pieces of mail necessary to vote which thus put the hammer to the most important progressive demographic.  This year, it’s looking like the vote will be in late October at the earliest (rather than early-to-mid October) which might help out the youth vote and make the discrepancy closer to that witnessed in 2006.  I could, of course, be totally wrong about this, but that is my feeling talking with staffers on Parliament Hill from multiple parties.

I certainly wouldn’t want to wager a great sum of money on it, but I would argue that this analysis still nevertheless establishes that Elizabeth May does have at least a fighting chance of securing the riding.


28 Responses to “Does May have a chance in Saanich — Gulf Islands?”

  1. 1 Robert McClelland 1 September, 2009 at 8:22 pm

    Although it’s far less scientific than what you presented, I think we can add in a pity factor that sees Lib and NDP voters casting their ballot for May in an effort to give the growing number of Green voters representation in Parliament. I think I’d do that if the NDP didn’t have a chance and May were running in my riding.

  2. 2 Bob Smith 2 September, 2009 at 7:50 am

    You forgot to take into account the fact that a large part of May’s “success” in the 2008 election was due to the gracious (if inexplicable) decision of the Grits not to run a candidate against her in Central Nova. If you compare her vote total as a green candidate in 2008 with the combined totals of the Liberals and the Greens in 2006, there was only a modest increase in votes resulting from her candidacy (from roughly 11,000 to 12,600, our roughly 5% of the vote).

    Accepting you methodology, once you factor this in (and assuming that the Grits will not gift wrap a riding for her again – a safe assumption), you get the result that, where Liz May runs in a riding her support goes up 12.5%, not 25% (since, in reality her increase in Central Nova wasn’t a 12,000 vote increase from the Green totals in the previous election, but a 1,600 vote increase from the combined Grit and Green totals in the previous election, a increase by 5% of the popular vote, not 30%). Even if you add in the more favourable environment in BC, she’s nowhere close to knocking off Lunn (though your numbers suggest that she may keep the Grits from doing so).

    Also, not to knock your calculations, but both your adjusted vote totals result in a higher number of voters than the raw projected votes (roughly 78,00 voters in the geometric projection vs. 70,000 in the raw vote projection). Presumably, in all three computations, the vote totals (though not the distribution) should be the same, since they’re all purporting to predict the same result with different methodologies.

  3. 3 paulitics 2 September, 2009 at 4:19 pm

    “You forgot to take into account the fact that a large part of May’s “success” in the 2008 election was due to the gracious (if inexplicable) decision of the Grits not to run a candidate against her in Central Nova.”

    I think this is a fair point. That’s precisely why I included in my analysis May’s run in London North Centre which was not only against a Liberal, but was against a strong Liberal candidate (who ultimately won).

    Are you saying that this somehow shouldn’t count?

    “in reality her increase in Central Nova wasn’t a 12,000 vote increase from the Green totals in the previous election, but a 1,600 vote increase from the combined Grit and Green totals in the previous election.”

    Are you assuming that 100% of Liberal votes migrated over to Elizabeth May? The only responsible way of calculating the matter is to do what I’ve done which is to look at the raw increases and decreases without including in calculations extremely tenuous assumptions such as that Elizabeth May simply secured 100% of the Liberal vote, 0% of the NDP vote, 0% of the Conservative vote. Furthermore, in taking the raw vote totals as you are doing, you risk discounting that the last election had an overall lower voter turnout than in previous elections. In my opinion, this makes such raw calculations not entirely useful.

    “Also, not to knock your calculations, but both your adjusted vote totals result in a higher number of voters than the raw projected votes (roughly 78,00 voters in the geometric projection vs. 70,000 in the raw vote projection). Presumably, in all three computations, the vote totals (though not the distribution) should be the same.”

    Unfortunately this last sentence is entirely inaccurate. Now, keeping in mind the caveat that I said specifically that I don’t vouch for the exact vote totals in my post, it is demonstrably untrue that applying arithmetic sums and geometric products to vote totals would equal the same vote total. If you read what I’ve written, you will see that I’m applying unequal products and sums to uneven numbers. This will almost always necessarily result in a different vote total. In fact it would be MORE suspicious if it equalled the same vote total as that would be evidence of the “numbers being made to fit the case” rather than the other way around.

    If you think my math is intended to skew the data in favour of May, I can only tell you that: 1) I’m not now, nor have I ever been a Green Party supporter. And 2) you are welcome to go over my math and see if I’ve made an error. It’s possible I made a mistake, but I used this exact same methodology in predicting the London North Centre by-election and I predicted the results accurately and I did it 2 months before any other polling firm had come to the same conclusion.

  4. 4 Greg Morrow 2 September, 2009 at 4:38 pm

    With all due respect, Paul, but there are so many problems with this analysis, I don’t know where to begin.

    I guess I would begin by saying the underlying premise of your analysis is wrong.

    You cannot take a simple average the net change in GPC support when EM ran in a by-election and a riding with no Liberal candidate. Both cases are not comparable to running in a general election with all parties running candidates.

    As you (should) know, a much higher % of votes in a by-election are personal votes than in a general election. In a general election the determining factor in 80-85% is the party (about 60-65% by party and 15-20% by party leader), while only 15-20% is the local candidate. In a by-election as many as half the votes can be personal votes. That’s understandable since in a general election the government is on the line, while it is not in a by-election. If EM was running in a SGI by-election, then LNC could be a comparable and she would do much better than she will in the general.

    The inclusion of Central Nova is even more problematic. The Liberals typically get 25% of the vote in CN, but didn’t have a candidate in 2008. I estimate that May won about 1/2 of the Liberal vote, another 1/4 stayed home and another 1/4 scattered between the CPC and NDP. But the absence of a Liberal also had the effect of collapsing the NDP vote by about 40%, since EM was viewed as the chief rival to Mackay. She would have pulled some of that even if there was a Liberal, but not 40%.

    And you don’t even account for the differences in resources. The campaigns in LNC and CN prior to EM spent literally nothing and had almost no volunteers. By contrast, LNC and CN were fully-funded campaigns with significant volunteer help both inside and outside the riding. SGI is starting from a much higher starting point than either LNC and CN were prior to EM’s arrival. Andrew Lewis was one of the GPC’s stronger candidates. His 2008 return is not yet posted on the EC site, but I know the GPC loaned it a significant chunk of money. In 2006, they spent about $20k (they got 10%) and in 2004 they spent $80k (they got 17%); I think 2008 was around $40k (again got 10%). It should be clear that, apart from the aforementioned problems of using a by-election and a riding with no Liberal, it’s also folly to believe that going from zero to fully-funded (as in LNC and CN) and going from $40k to fully-funded will have the same effect, which is what you say when you simply take the average change in support and apply it to SGI (again, beyond the problems of even using LNC and CN as comparables).

    But then you make it even worse by making an adjustment for the change in party support in Atlantic Canada with respect to CN, but don’t do the same in Ontario with respect to LNC.

    Worse, you don’t make any adjustments for changes in support for the other parties in either LNC or CN.

    Note how the + and – don’t add up.

    And you cannot assume that the change in party support in SGI will precisely mirror the changes in party support in BC.

    And your baseline poll numbers are quite correct either. The Conservatives are higher than 33.9% and the Greens are lower than 14.5%.

    And even when you’ve got the right polling numbers, you have to adjust for the difference between polling and ballot box support. The Greens lost about 4 points at the ballot box in BC, beyond their final polling average in 2008. This is understandable for many reasons — last minute shifts, relative effectiveness of GOTV operations, strategic voting, and so on.

    By contrast, the CPC went up by about 6 points beyond the final polling average in BC. These are fairly well-established patterns — the Conservatives being under-estimated in BC polls, and the Greens being over-estimated.

    There are also problems with your total # of voters as noted in the comment above. I think the point being made above is that turnout is turnout, it doesn’t change because you use different ways of crunching the numbers. Moreover, the total number of votes is far from reality. In one case, you have over 80,000 votes. Considering there are only 95,000 voters, that would be a turnout of almost 85%. Turnout will be closer to 70%, closer to 65,000 votes.

    I could go on, but you get the point. In short, and again with all due respect to your effort, your analysis has so many holes, it just doesn’t hold water. Obviously, you won’t agree with that assessment, so I was trying to think how you could make adjustments, but I’m not sure it can be salvaged because the underlying premise of using LNC and CN as your baseline is flawed.

  5. 5 paulitics 2 September, 2009 at 7:24 pm

    Hi Greg,

    Thanks for your reply.

    First off, in your lengthy reply, you did make a very important point. Yes, I did forget to take into account the (considerable) amount that the Greens tend to drop from their polling projections to their vote totals. I’ll be updating this post shortly to account for that.

    My initial paper napkin calculations still give May at least a fighting chance of winning the riding even with this revision, however, I’ll withhold judgement on this until later.

    Second, you wrote:

    “In a general election the determining factor in 80-85% is the party (about 60-65% by party and 15-20% by party leader)”

    This is absolutely true for most political parties and in most political circumstances (and there are numerous peer-reviewed studies that establish this). However, when Elizabeth May is running (and it’s been demonstrated twice now) her riding becomes a de facto by-election even if the race takes place during an election. A riding with Elizabeth May is not, to put it bluntly, a normal riding.

    By your logic then, there is absolutely no explaining how May was able to do so much better than her Green counterparts in Central Nova if “80-85%” of her support was already locked up.

    So I’m going to have to say that your statement there really needs some serious support because from where I’m standing, it’s not tenable.

    Third, you then write:

    “But then you make it even worse by making an adjustment for the change in party support in Atlantic Canada with respect to CN, but don’t do the same in Ontario with respect to LNC.”

    You actually give the reason for this yourself. You accurately stated earlier that by-elections are not usually (as) continent upon polls but rather on the individual candidates (take, for example, Elizabeth May in London North Centre or Thomas Mulcair in Outremont). So I think you’re attributing a great deal of ignorance to me here without considering that there may be a very good reason for doing this the way I decided to do it.

    Fourth, you write:

    “And you cannot assume that the change in party support in SGI will precisely mirror the changes in party support in BC.”

    I fully agree with this. This is a huge problem with the projection model as I’ve done it. I don’t think anybody is under any illusions that provincial polls actually translate perfectly down to each riding level. Of course they don’t. But polls are still, for better or worse,the best data points we have when it comes to predicting these kinds of things.

    Fifth, you write:

    “Note how the + and – don’t add up.”

    I did notice this. You must really think I’m profoundly incompetent not to recognize this. I don’t think that this is a particularly problematic issue. To show you why this isn’t a problematic issue, just take a look at the 2006 results for the Ottawa South riding and then look at the 2008 results.

    ——- 2006 ———–
    Liberal 27,158
    Conservative 23,028
    New Democrat 8,138
    Green 2,913

    ——- 2008 ———–
    Liberal 29,035
    Conservative 19,417
    New Democrat 4,920
    Green 3,939

    Now look at how much each party went up or down in 2008 from their 2006 results:

    (geometric change individual party’s vote)
    Liberal +6.9%
    Conservative -15.7%
    New Democrat -39.5%
    Green +35.2%
    TOTAL: -13.1%

    So by your argument, there’s something wrong with the 2008 election results because the sum totals of the geometric change in each individual party’s vote total don’t add up to 0?

    Fifth, you write:

    “And your baseline poll numbers are quite correct either. The Conservatives are higher than 33.9% and the Greens are lower than 14.5%.”

    I know that the Greens are not that high nationally and my polling data doesn’t say anything of the sort.

    My data is looking only at British Columbia polling numbers (rolling five poll averages to be precise) not at national numbers. You really are judging me quite harshly without fully understanding where the data is coming from.

    If you’re interested, my provincial polling data is both accurate and up to date and you can access it here to see where I got these numbers:

    You can also verify my numbers by going here and seeing for yourself that my data is, in fact, correct.

    Finally, you write:

    “Moreover, the total number of votes is far from reality. In one case, you have over 80,000 votes. Considering there are only 95,000 voters, that would be a turnout of almost 85%. Turnout will be closer to 70%, closer to 65,000 votes.”

    I don’t believe for one second that 80,000 votes will be cast in this riding. This is, incidentally, why I wrote (in English no less) that:

    “Now, I do not vouch for the exact vote totals appearing in these final two columns. However, as a rough predictor of how probable an Elizabeth May victory is, I think this is a strong indicator that it is entirely possible.”

    The key thing I’m looking at is the relative percentages of each party not the overall vote total.

    I don’t think you’ve taken the time to read through my argument in good faith.

    • 6 Bob Smith 2 September, 2009 at 9:39 pm

      Paul, I’m taking your reasoning seriously, but I think you have to take into account some of these factors.

      I mean the fact that the Grits didn’t run a candidate against her (and actually endorsed her) had to have helped her performance in Central Nova. To suggest otherwise is ludicrous. That doesn’t mean all the grits voted for her, but it’s equally ludicrous to assume, as you did, that none of them did. Clearly many of the people who voted Liberal in 2006 voted Green in 2008,that’s the only way Liz’s numbers work out. But would they have voted for May if they’d had a choice?

      And this point also reinforces Greg’s point about May’s support in SGI being locked up. Why did the Greens do so much better in Central Nova in 2008 then in 2006? Surely the fact that they were endorsed by the official opposition had something to do with it. There was room for growth there because the Grits had vacated the field and left a large pool of potential voters open to the greens. But that was such an exceptional set of circumstances, it’s not reasonable to expect it to be repeated.

      As for the total number of votes, I think it matters because it’s what changes (at least in part) the percentages of each party. You’re getting different percentages because you’re adding in hypothetical (and non-existent) voters. I take your point that at the end of the day it’s the percentages that matter, but if you can’t change the percentages while holding the total number of voters constant, there’s a problem. When you add an extra 10,000 voters to the pool and they all seem to vote for May (which is the result you have in geometric projection), there’s a problem with the formula. Unless you’re assuming that Liz May will stimulate voter turnout (which doesn’t appear to have been what happened in Central Nova – London was a by-election so you can’t read much into turnout there, though for the same reason I also wouldn’t read much into the result), you need to tweak that computation.

  6. 7 Bob Smith 3 September, 2009 at 7:56 am

    Just as a follow up, I now realize what’s bothering me about your prediction. You’re counting some voters twice. Basically, as I understand it, you’re making your prediction by taking the previous results and then adding (or subtracting) votes to reflect two effects (1) the increase (or decrease) in votes caused by Liz May being a candidate in a riding (although I have some doubts about the magnitude of that effect, there’s clearly some effect) and (2) the increase or decrease in votes caused by the change in popularity of the party.

    The problem is that your methodology doesnt’ account for the possibility that the people who change their vote because May is a candidate might also be the people who would otherwise change their vote because, for example, the Green’s are more popular (I suppose it would be more accurate to say they’re the people who have changed their votes in a manner reflected in the polls). You’re assuming that those are two separate pools of people, but they’re not, presumably people who might be inclined to change their vote to green are also those people who might be inclined to change their vote to vote for May. That’s why in your projections you end up with too many voters.

    The problem is apparent with the Greens (because of the size of your suggested May effect and the party popularity effect and the fact that they work in the same direction). Your predictions for the Tories and NDP are probably somewhat more accurate (because the size of the May effect and the Party popularity effect, respectively, are relatively small, so the extent of double counting is minimal). Your prediction is probably also pretty close for the Grits (because the May Effect and the Party Popularity effect have opposite signs, they tend to cancel one another out). That’s why you end up with the weird result in the Geometric prediction where you have an extra 10k voters (vs. the raw numbers), who seem to all vote for the Greens. The Greens are the only party, in your methodology where this double counting leads to a significant difference.

    Unfortunately, I can’t think of a rigorous way of correcting for this double counting. But as a back of the envelope starting point, I’d say that the NDP, Liberal and Tory projection numbers are probably pretty close to reality (because, as I said, I think the degree of double counting for those parties is pretty small) and, because I think the May effect wiill swamp the party popularity effect for the Greens (in that people who would change their vote to the greens anyhow are probably those people who would vote for the greens because May is running), the Green’s raw numbers are probably a good prediction of how they’ll perform (based on what I think are your debatable conclusions about the side of the May effect). On your original data that produces a pretty close three way race between the Tories, Grits and Greens (though, if you make the subsequent adjustments Greg Morrow suggested, I think you end up with the Tories out ahead).

  7. 8 Dave Bagler 3 September, 2009 at 8:56 am

    Hi Paul,

    I just wanted to thank you for posting your reply to Greg Morrow on my blog as well.

    You’re right he didn’t go through your argument in good faith. Had you used the same reasoning and came to the conclusion that Elizabeth would lose, Greg would have applauded your analysis.

  8. 9 Brian Gordon 3 September, 2009 at 12:42 pm

    Interesting analysis. As a former Green Party of Canada candidate in Esquimalt-Juan de Fuca and as someone deeply concerned about Canada’s non-response to climate change, I will be following this race closely.

    There are some wildcards that, depending upon how played, will influence the outcome of the SGI vote:

    * Elizabeth May was accused by people within the GPC and generally of supporting strategic voting during the last federal election. In fact, this split the SGI riding, and will certainly be thrown in her face by opponents. If strategic voting made sense then, why not now? As I pointed out then, Anything But Conservative is insulting and, in reality, translates practically to Anyone But Green.

    * Many Greens not thrilled with Ms. May’s leadership have left the party, some noisily. (See: David Chernushenko, for example)

    * EMay is associated with Dion’s Green Shift, which scared many conservatives in BC (and elsewhere). Same goes for perceptions of Green economic policy – which is unfortunate, because it makes more sense than the business-as-usual that the other parties offer. Nonetheless, the Greens have not done well at getting the benefits of the green economy across, and she’s running in a riding that has sent a Conservative to Ottawa more than once despite his screw-ups.

    * Will EMay get into the leader’s debates again?

    * The Liberal candidate last time around was a former Green; who will the Liberals put up against EMay this time?

  9. 10 Steve 3 September, 2009 at 12:43 pm

    Hi Paul,

    Interesting analysis, but I think it is more complex than this, and honestly, as a GPC supporter and former resident of SGI, I don’t think EMay has a chance. Look at the GPC %vote total of both LNC and CN prior to EMay – 1.7% and 5.5%. 5.5% is about average for GPC, but 1.7% is a low result. On the other hand, SGI has had consistently higher than average numbers for the GPC (10.45, 9.94, 16.71 for the last 3 federal elections respectively), but this may mean that it is harder to increase the numbers much more. Further, your prediction assumes a drop of NDP vote by more than 50%, down to roughly 12%. However, even with the candidate pulled out, 5.69% of SGI voters still voted for NDP. This is a complex riding, one worth watching in every election in the last decade, and one in which both the NDP and Liberal candidates in the 2008 election were former Greens. The liberal candidate especially has strong green credentials and has strong public name recognition from her 3 years of hosting Enviro/Mental on CHUM. Thus, the small ‘g’ green vote gets split 3 ways in this riding, the same way some see the “left vote” as being split.

  10. 11 SIR 3 September, 2009 at 1:54 pm

    Bizarre analysis. Essentially, you are suggesting Gary Lunn’s support will collapse down to 27% from last time (42%?) largely because Elizabeth May runs in SGI?

    Have you looked at her positions? They are off the left end of the spectrum. And the Lib candidate has even better bob\nafides thatn her (Western MBA , environmentalist, well known in the riding).

    A case of one who can’t see the forest for the excel spreadsheet.

  11. 12 Greg Morrow 3 September, 2009 at 2:45 pm

    @Dave: I’m not critiquing the conclusion per se (although obviously once you fix the methodological problems I’ve outlined, you would get a different result). You seem to suggest I have a preference where EM runs (or worse, that I want to see her lose, which isn’t the case). I don’t have a preference whatsoever, she should run in the riding that gives her the best chance of winning. All I’ve said is SGI was not that riding (but I understand why she cannot run in the best riding, Guelph — because of tensions with the local team and because she won’t run against a Liberal).


    @Paul: actually, I read your analysis quite carefully, but I think you aren’t quite understanding a few of my points, or perhaps I didn’t explain them well enough. Let me address your latest comments.

    To reiterate my key point: you’ve made some basic false assumptions because you believe that the previous two times EM has run are predictive of what will happen in SGI.

    You will not see the same increase in GPC support in SGI as you did in CN or LNC for several reasons: (1) there was no Liberal candidate in CN, thus the change in support for the GPC was highly inflated, (2) about 80% of votes in general elections are party votes vs only about 50% in by-elections, so more of the increase we saw in LNC is attributable to personal votes than will be in a general election in SGI, and (3) the campaigns in LNC and CN had no resources prior to EM’s arrival, while Saanich spend $40k last time, so it is untenable to think that going from zero (and no volunteers) to $80-90k (and many volunteers) will have the same effect as going from $40k (with volunteers) to $80-90k. In all 3 cases, because you didn’t account for these factors, you are seriously overestimating the expected change in GPC support in SGI.

    You’ve simply not addressed these points. What you are saying is that it doesn’t matter if there was a Liberal or not, EM will see the same increase in vote (on avg) in SGI as we saw in CN. You are saying it doesn’t matter that more votes will be party votes in a general, EM will see the same increase in vote (on avg) in SGI as in LNC. You are saying it doesn’t matter that LNC and CN were starting from zero while SGI spent $40k last time, she will see the same increase in vote from 2008. These assumptions are simply not plausible.

    Had you understood my points, you would not have said “by your logic then, there is absolutely no explaining how May was able to do so much better than her Green counterparts in Central Nova…” Indeed, I did explain it: since there was no Liberal candidate, EM won 1/2 the previous Liberal vote about 40% of the NDP vote. In other words, a good portion of the votes she received was DUE to the absence of a Liberal candidate. Likewise, EM had a fully-funded campaign vs essentially nothing previously. I’m not saying she didn’t deliver personal votes — she did, and that can be estimated, and she will do so again in SGI, as I outlined in my analysis on DS — but the % of votes attributable to EM (i.e. personal votes) is not nearly as high as you think it is, and certainly not the 100% as you’ve assumed by not accounting for the above.

    Likewise, more % of votes in a general election are party votes regardless of the local candidate (although a bit higher when a star candidate is running), as compared to a by-election when the % of party votes is lower. So assuming she will attract the same number of personal votes in SGI in a general as she did in LNC by-election is just not supported by what we know about the differences in voting behaviour between by-elections and generals. That is, your assumption that SGI will be like a “de-facto by-election” just cannot be supported. In fact, we know there is a difference — it’s the difference between party votes (80% vs 50%).

    To make your analysis hold water, you need to make significant downward adjustments to account for the absence of an LPC candidate (in CN), the difference between personal and party votes (in LNC), and the different starting points between SGI ($40k spent + volunteers) and CN/LNC (nothing).

    Let me explain more carefully my point about not making adjustments for shifts in party support in LNC whatsoever and only making the adjustment for the GPC in CN. By not making adjustments for changes in party support in LNC, you are attributing 100% of the vote to personal votes. But that’s not the case — it is, as I noted, much higher in by-elections, but it is not 100%. It’s more like 50%. In other words, for example, 1/2 the decrease in the Liberal vote can be attributed to the decrease of Liberal support in Ontario from the 2006 election up to the by-election. You need to make that adjustment and multiple it by 50% (as opposed to 80% you should normally do for general elections). You simply have not accounted for that change in party support. Likewise, by not making adjustments for changes in party support in CN, you are attributing 100% of the changes to EM. So you DO need to make these adjustments in both CN and LNC for all parties for your analysis to hold water.

    About my point re the +/- not adding up — the change in party support from 2008 to now is NOT a geometric calculation (i.e. % change from last time), it is a NET change in support (in points). You are comparing apples and oranges, and confusing the two.

    So to use your Ottawa South comparison, the NET change in support for the Liberals is not +6.9% (that is the geometric % change over the previous result i.e. 29035-27158/27158); the NET change was actually +5.7 points — David McGuinty went from 44.2% to 49.9%. NET change in support always evens out; a % change will not.

    And note that you are indeed using a NET change in support when calculating the average in CN and LNC. It is not a % change over the previous result; if it was the % change in CN would have been 1780% — i.e. 12620-671/671. You are confusing % change (what you use in your Ottawa South example) vs NET change (what you use in your average change in support). Likewise your adjustments later are NET change in support — you have CPC -10.5 (i.e. 44.4 – 33.9 = 10.5), so they are clearly NET change not % change. % change is obviously meaningless as the 1780% attests. So you should be using NET change and if so, the numbers should even out to 0 (i.e. the + should equal the -), but they don’t because you are averaging two ridings, which should be a clue that the method isn’t that sound.

    re: your polling data, it is more prudent to use the average of the last poll for each of the 5 major pollsters (Ekos, Nanos, Harris-Decima, Angus Reid, Ipsos), than to simply use the last 5 polls published. Doing the latter means that 3 of the 5 polls are Ekos, since they publish every week. In other words, your average mostly reflects what Ekos polls show, not the cross-section of all pollsters, which is a more reliable metric. And I noticed that you don’t have ANY Ipsos data, which is also a problem.

    And, you also need to recognize when you have an outlier. It may be OK to include the bizarre Harris-Decima poll that shows the GPC at 24%, but when the other 4 pollsters latest show an average of 12.3%, literally half as much, then it is clearly an outlier. Also, just a bit of history from last election: Harris-Decima’s last BC poll, just 4 days before e-day, had the GPC at 21% (they got 9.4%). The prudent thing is to treat the 24% as an outlier. For the GPC, the other pollsters currently have: Ekos 14.3, Angus Reid 10, Ipsos 14, Nanos 11.1 = 12.3 avg (vs 14.9 you have). Likewise, for the CPC: Ekos 35.3, Angus Reid 43, Ipsos 38, Nanos 32.3 = 37.2 avg (vs 33.8 you have). Note also that the Nanos poll is older, so is now likely higher for the CPC, though not necessarily lower for the GPC.

    And as to your final point, all I’m saying is that raw numbers and turnout matter. It is simply not possible for EM to win 25,000 let alone 32,000 votes. Take a look at my post at DS that breaks down SGI’s demographics, hopefully it’s clear why this is the case.

  12. 13 marcel 5 September, 2009 at 4:03 pm

    that’s the longest reply of all time on this blog! here’s you’re hero cookie


  13. 15 Hobbes 25 January, 2010 at 12:32 pm


    I’m starting a new Libertarian Socialist blog called Rosa’s Ghost. I
    was hoping you might add it to your links page. It would be a great
    help. Thanks in advance.

  14. 16 marcel 24 February, 2010 at 1:10 am

    Ain’t no blog there, and Paulitics has disappeared too. As you may have noticed. I sure did, come back Paulitics, come back.

  15. 17 Cayer 25 February, 2010 at 12:12 am

    Yeah! This was a good blog, come back!

  16. 18 Joshua T. 22 October, 2010 at 10:53 pm

    Excellent post. BTW, here’s a link to the Straight Goods

  17. 19 crysis ati patch 25 February, 2013 at 5:11 pm

    Therefore, the characters must sell animal skins in order to
    get money for buying the weapons. I even find myself backing away from recommending other resources which, in
    the past, I would have shared willingly. I thought it would be cool to make people
    laugh as I inquire into the things that fascinate
    me, like the trap of being human and, at the same
    time, divine. Fun the first time, but just not enjoyable after numerous journeys.

    In doing so, it becomes the only potential choice for best gaming PC under 500 bucks in the desktop category.
    For an in depth guide to the E8400 chip, I highly recommend reading this article: Final
    Thoughts on the Intel E8400. We relate to it like it’s the truth. I’m the
    kind who smokes and judges people’ Vanda pretends to demonstrate a yoga pose with a cigarette in her hand’ ‘Whoa, THAT guy’s out of shape”.

    *Composite fillings. Therefore, you cannot afford to miss this incredible shooter game involving a series of killings.

  18. 20 4 March, 2013 at 6:38 am

    The fact that Toyota chose to upgrade the exterior styling, interior quality and interior roominess of the redesigned 2012 Yaris
    hatchback rather than upgrading the powertrain says a lot about how Toyota views the needs of small car buyers.
    Another thing you’ll need is some strength to endure, because there’s a
    good chance you’ll be your own producer for the show or perhaps the best producer the show has. Even today, I go big game hunting and all this, and the more dangerous something is, the happier I am.

  19. 21 fat loss factor review 25 July, 2013 at 8:39 am

    Hi, You must have done a fantastic job. I am going to certainly bing this and professionally advise to be able to my pals. Almost certainly they will be took advantage of this excellent website.

  20. 22 iphone 5 case 12 September, 2014 at 4:41 am

    So, how can one be a musician through their car stereo.
    This time apple was able to pack 4 times the amount of pixels into the phone compared to the previous versions.
    A close examination of the i – Pad’s features could answer this question. which pretty much rules out sexting, parents will be
    relieved but seriously, a truly portable communication device should have included a webcam.

  1. 1 Waiting for the next Canadian federal election - Counterweights Trackback on 11 September, 2009 at 11:02 am
  2. 2 Canadian federal election update : September 11 - Counterweights Trackback on 11 September, 2009 at 1:46 pm
  3. 3 Corey Park scam Trackback on 4 March, 2015 at 6:48 pm
  4. 4 Foundation for Defense of Democracies Trackback on 6 March, 2015 at 2:25 am
  5. 5 Ian Leaf Trackback on 13 March, 2015 at 10:10 pm
  6. 6 Trackback on 6 October, 2015 at 6:40 am

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


home page polling resource

Click below to download the

Paulitics Blog Search

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 2.5 Canada License.

DISCLAIMER: The views and opinions expressed in the comments section beneath each post on this blog do not necessarily reflect the views and opinions of the blog's author and creator. Individual commentators on this blog accept full responsibility for any and all utterances.


Progressive Bloggers

Blogging Canadians

Blogging Change

Paulitics Blog Stats

  • 864,177 hits since 20 November, 2006

%d bloggers like this: