Monday, April 09, 2012

Mulcair's win was inevitable...but was it predictable?


PunditsGuide concludes her top notch NDP leadership coverage with the vote breakdown of advance and in-person ballots. White Nathan Cullen "won" the convention vote, Mulcair had enough support among advanced voters that he could have dropped his pants and sung La Marseillaise during his Friday night speech, and likely still carried the day.

So the question is - if a Mulcair win was inevitable, was it predictable?

I compiled a series of 9 leadership metrics in the days leading up to the convention on attributes such as fundraising, endorsements, social media, and polling data. In my mind, this was a bit of a test drive for the Liberal leadership race, since I'd like to be able to post weekly or monthly "Power Rankings" during that contest, to gauge candidate support and momentum.

So here's how the different indicators stacked up, sorted by their correlation to the first ballot results:

Poll: 0.899
Endorsement Score: 0.869
Media mentions: 0.868
MP Endorsements: 0.860
Donations: 0.836
Twitter mentions: 0.715
Donors: 0.683
Facebook "likes": 0.308
Twitter followers: 0.170

As has been the case in past leadership contests, polls among party members proved to be the most accurate prediction of candidate support. And remember, in this case we were dealing with a pair of month-old polls conducted by two of the leadership camps. Of the two, Mulcair's proved to be more accurate - mainly because (surprise, surprise) Paul Dewar's overstated his support by a factor of 2. Let that serve as a cautionary tale anytime you see an "internal" campaign poll released.

Endorsements also score well, using either a simple count of MPs, or 308.com's more complex system. Mind you, Premier Mar and Premier Falcon will tell you how much endorsements can be worth in some contests.

Fundraising numbers were a good indicator of candidate support, though not as much as the 2003 NDP leadership contest when they had a remarkable correlation of 0.998. As has been the case with the other half dozen leadership races I've crunched the numbers on, the total amount raised seems to be a better predictor than the total number of donors. Goes to show money matters, even when we're talking about socialists.

As for social media? It may play a big role in modern elections, but it's certainly not a very good indicator of who's in front. Language may have played a role in it, but Mulcair was 5th in terms of Facebook "likes" and was barely ahead of Niki Ashton when it came to Twitter followers.

Surprisingly enough, something as simple as the number of times a candidate is mentioned by the media proved to be just as good a predictor of support as the factors above. So while we all like to think the media has no clue what goes on during internal party contests, they were basically on the mark this time. Goes to show you there's some truth behind what's heard at Hy's.

If we toss all these numbers into the regression wood hopper, the best prediction models weights endorsements, donors, polling, and media buzz about evenly, and explains 92% of the variance. That's a good number, but my stats teachers would slap me if they saw me creating a "magic regression equation" based on 8 candidates in one leadership race. After all, each contest has its own rules and personality, and it's foolish to think the Liberal race will look like the NDP's.

Still, the signs that mattered most all pointed to a significant first ballot lead for Mulcair. We shouldn't be at all surprised that's what happened.

Labels: ,

7 Comments:

  • If the convention was in BC instead of Toronto would Cullen have been able to edge out Topp and possibly win? His support base was supposedly in his home province but he managed to get a lot of people to the convention to support him, I doubt there were a large number of BCers there.

    By Blogger Jordan, at 1:26 PM  

  • Thanks for the props, Grit.

    FWIW, a number of observers of the NDP race felt that Dewar's poll was a decent reflection of the state of the race at the time, i.e., at the beginning of the year. Topp's folks didn't, but they didn't like the results.

    Then the french debates changed that, the new members came on board, and the Broadbent remarks worked to revive Topp's candidacy at the end.

    I do believe Dewar's camp waited for the results they wanted before releasing them, but I don't believe they cooked them.

    The Jurist has a pretty good post on candidate what-ifs, in case you didn't see it, where he considers the decision to release the poll, decisions about how to react to it, and its impact on the entire race.

    http://accidentaldeliberations.blogspot.ca/2012/04/leadership-2012-candidate-what-ifs.html

    Anyways, thanks as well for your power rankings, and keep up the great work.

    -------------

    And to Jordan: because the voting was online, rather than by convention delegates, it seems the major impact on Cullen was the presence or absence of a big floor demonstration. Nash seemed to be the beneficiary of a Toronto convention location in that regard, not that it helped her on the vote. Don't forget Topp was also strong in BC.

    By Blogger The Pundits' Guide, at 1:34 PM  

  • Money. All races have different correlation, but, donations always show a respectable result.

    By Blogger ridingBYriding, at 2:18 PM  

  • PunditsGuide - I wouldn't expect Dewar too cook the results, but campaigns clearly wait until the "best" numbers before releasing them. I'm not sure what Dewar did, but if I was doing a weekly rolling poll of 1,000 delegates, I'd certainly wait until I got the best one possible (be it due to real momentum or statistical variation in the numbers) before releasing it.

    The fact that Mulcair's numbers also showed Dewar doing respectably (if not as well), suggest he actually was in decent shape at that point.

    By Blogger calgarygrit, at 2:28 PM  

  • The NDP 2008 platform to use cap revenues for green R+D could be tweaked to get more from less. The safest and cheapest (geothermal might not get cheap everywhere, may never get CNT electrical lines, would be a good NNI try in Edm) way to solve our climbing emissions is developing utility grid scale battery banking of intermittent wind and solar sources. Right now these demoes are swimming pools of chemicals. In the future they might resemble PEM fuel cells in appearance and coarse function, or they might still be giant vats of chemicals.

    The investment in Ballard only gave forklift batteries instead of desired vehicles. I'd try again with batteries. At its core raising corporate taxes is the best way to diversify and there are many other industries...Sony laid off a chunk of their workforce and they are very employment intensive. We outta be developing hologram teaching curriculum and technologies...lead this telecom wave. Be good in pandemics, for up north, for our universities, and the world will copy. Big batteries and holograms over bank dividends and oil profits (wind batteries and CNT can be AB specialties, CNTs use methane in CVD furnace), I think is the most efficient industrial strategy. Flames did same thing Jets did, not getting replacement players for injuries. Leafs...all downhill since that Gretz bankshot.

    By Anonymous The Keystone Garter, at 2:32 PM  

  • Given the number of early voters, and the tiny number of online votes/undecided delegates in play, it probably could have been called after the first vote if the numbers had all been known. For example, of the 4,500 delegates actually at the convention, only 1,500 hadn't already voted (so the MSM says). So the convention gave the illusion of drama, but basically the game had already been decided.

    By Blogger bigcitylib, at 6:52 PM  

  • "Surprisingly enough ... the number of times a candidate is mentioned by the media proved to be just as good a predictor of support as the factors above."

    Doubtless that's in part because mentions in the media helped cause the outcome.

    By Blogger Brian Henry, at 5:23 PM  

Post a Comment

Links to this post:

Create a Link

<< Home