Warning: include(new/inc/header_ads.php) [function.include]: failed to open stream: No such file or directory in /home/punditsg/public_html/wp-content/themes/et-starter/single.php on line 194

Warning: include(new/inc/header_ads.php) [function.include]: failed to open stream: No such file or directory in /home/punditsg/public_html/wp-content/themes/et-starter/single.php on line 194

Warning: include() [function.include]: Failed opening 'new/inc/header_ads.php' for inclusion (include_path='.:/usr/lib/php:/usr/local/lib/php') in /home/punditsg/public_html/wp-content/themes/et-starter/single.php on line 194
  ››  Pundits' Guide

More 2011 Election Metrics – Department of Cautionary Tales

May 9th, 2011

[Welcome, National Newswatch readers!]

Continuing our series of post-election metrics, I wonder: could we please lay the following faulty rules of thumb to rest, once and for all.

1. If a riding was close last time, it's a riding to watch this time.

No. In fact this is now the Pundits' Guide Prime Directive™:

"Just because a riding was close last time, doesn't necessarily mean it will be close this time."

Of the 42 ridings that were close in the last general election (i.e., within 5%), only 12 were still close this time. And 39 of the ridings that WERE close this time – that's the vast majority of them – were NOT close last time.

    2008 General Election
  Marg <= 5% Was Close Not Close ALL
2011 GE Was Close 12 39 51
Not Close 30 227 257
  ALL 42 266 308

Or, if you prefer the criterion to be within 10%, only 30 of the 79 races that were within 10% last time were still within 10% this time. And 51 of the close races in 2011 – again the vast majority of 81 closest contests – were NOT close contests last time.

    2008 General Election
  Marg <= 10% Was Close Not Close ALL
2011 GE Was Close 30 51 81
Not Close 49 178 227
  ALL 79 229 308

By the way, four ridings have been within 5% in each of the last four elections:

  • Ahuntsic, QC – (and who knows how this riding will look next time around, if the Bloc Québécois no longer exists as a federally registered party)
  • Burnaby-Douglas, BC
  • Esquimalt-Juan de Fuca, BC – (and if I had to guess, won't be close next time, since the Liberal anomaly here won't be a factor next time around)
  • Vancouver Island North, BC

2. We can know which seats are most likely to change hands this time (and will be "ridings to watch"), because they will have been amongst the closest seats last time.

In incremental elections, just under half the seats that will change hands were previously won by 5% or less, and around three-quarters will have been previously won by 10% or less.

But in transformational elections, those percentages drop substantially. In 2004 and 2011, just under one in five of the seats that would change hands had previously been won by 5% or less, and only around one in three had been previously won by less than 10%.

Thus, in order to know whether the previous margin is predictive of a seat's likeliness to change hands, it seems you need to know ahead of time whether the election will be incremental or transformational. In other words, good luck with that.

This methodology could miss at least a quarter – and as many as two-thirds – of all the ridings that will change hands in an election.

[UPDATE: I described the difference between incremental and transformation elections in a blogpost, found here, before the current election.]

Seats Changing Hands Party-Wise, by Previous Margin (either By-election or General Election), 2004-2011

Bin 2004 2006 2008 2011
0 – 5% 28 25 19 18
5 – 10% 16 18 10 19
10 – 15% 10 3 6 13
15 – 20% 19 4 3 13
20 – 25% 15 2 1 13
25 – 30% 9 3 1 7
30 – 35% 6 0 0 8
35 – 40% 5 0 0 4
40 – 45% 8 0 0 2
45 – 50% 7 0 0 0
50 – 55% 5 0 0 1
55 – 60% 5 0 0 0
60 – 65% 1 0 0 0
Rest 0 0 0 0
 
All 134 55 40 98
Pct <5% 20.9% 45.5% 47.5% 18.4%
Pct <10% 32.8% 78.2% 72.5% 37.8%

 

3. If a riding were going to change hands, the second-place finisher last time will usually be the party that wins it.

Only in incremental elections. Not the case in 1993, 2004, 2011.

Note that the government didn't change hands until 2006, but 2004 saw the most upheaval (the result of the merger between the two conservative parties).

Seats by Previous Rank of Winning Party (either By-election or General Election), 1993-2011

General
Election
Previous Rank of Winning Party Total
Seats
1 2 3 4 5 n/a*
* Winning party did not exist, or winner had been/became Independent, etc.
2011 GE 210 38 15 45     308
2008 GE 268 38 1   1   308
2006 GE 253 46 7 1   1 308
2004 GE 175 39 61 30 2 1 308
2000 GE 269 25 7       301
1997 GE 248 35 12 5   1 301
1993 GE 93 86 26 24 5 61 295

You'll notice that I picked these three assumptions, because they were the ones used by the various amateur seat projector/predictors and the strategic voting sites which depended on them. We'll be taking a look at those next.

17 Responses to “More 2011 Election Metrics – Department of Cautionary Tales”

  1. Immanuel says:

    Alice,

    Is there any way to calculate which parties were most able to convert 2nd, 3rd, and 4th places to victories ?

    Eg. how many 2nd place CPC in 2008 converted in 2011 or how many 4th place NDP converted in victories in 2011.

    And per province, if it’s possible.

    Just wondering.

    Immanuel

  2. Immanuel says:

    p.s.
    And how would each party score on the incumbency re-election index ?

    e.g.
    how many CPC incumbents were re-elected in 2008 and 2008 versus LIB and NDP.

  3. Immanuel, you may have me confused with a data drone. There are tools here at Pundits’ Guide that would let you count all of that up yourself, you know.

  4. Brendan Kane says:

    So outside of Quebec there were only 6 seats where a party came from 3rd place:Scarborough
    Southwest, Scarborough Rouge River, Esquimalt, Newton North Delta, Saanich-Gulf Islands & Labrador

  5. Yes, I can confirm that those were the six outside Quebec. The nine in Quebec were:

    Hull – Aylmer
    Jeanne-Le Ber
    La Pointe-de-l’Île
    Laurier – Sainte-Marie
    Montcalm
    Pierrefonds – Dollard
    Rosemont – La Petite-Patrie
    Saint-Hyacinthe – Bagot
    Verchères – Les Patriotes

    Now, I suspect there were a lot more seats where second and third place switched, but haven’t looked at that yet.

  6. Observant says:

    And here is how it all turned out in the wash:

    1. Conservatives won 161 ridings in the RoC and only 6 in Quebec. That means they won 161 of an available 233 risings in the RoC for a 69% MAJORITY and an overall 39% nationally.

    2. NDP won 44 ridings in the RoC and 58 in Quebec. Those numbers speak for themselves, but it also means if Pepsi Quebec had not gone bonkers for Orange Crush, the NDP may have only won perhaps 48 ridings max. This would mean the BQ would have been the OOP if fate had unfolded as it should have.

    3. Liberals with only 34 MPs may drop further if the ancient Liberal MPs decide to retire with their gold-plated parliamentary pensions …. resulting in by-elections and further Liberal losses.

    Historians are people who think backwards and statisticians who count backwards. Time to ponder the future possibilities.

    So, Alice … care to guess which of the old Liberals will retire and which way their ridings will go in by-elections??

    (Loved ya on CTV …!!!)

  7. Although, I should add in the next breath, Immanuel, that you have been of incredible help to me, and I shouldn’t be too huffy with you.

    I’ll see if I can look it all up for you. And sorry for being “the thing that rhymes with witch”.

  8. Mike says:

    I like your omment from #3: “Only in incremental elections. Not the case in 1993, 2004, 2011.”

    That’s interesting. No one knew at the start of 1993, 2004, or 2011 that those elections were going to be transformational and not simply incremental. In other words, there is a serious flaw in seat projections – one would need to guess in advance whether there is a major change afoot or whether we are only experiencing an incremental change.

  9. Shadow says:

    The whole issue of MPs retiring needs to be looked at.

    After they lost gov’t in ’06 a large number of Liberals basically quit going to work while still occupying their seat.

    Jack Layton took Michael Ignatieff to task for being absent from Ottawa during the debates.

    How about Conservatives and New Democrats work together to design a system of penalties for MPs who miss work ?

    Perhaps even make it so that if they don’t show up on a regular basis their seat is declared vacant.

  10. Alison says:

    Alice, if you have time, could you comment on George’s link above?

  11. Hi Alison, It’s probably worth a post on its own to examine the math. I’m still working on strategic voting sites, but will look at that next, if there’s an interest. Thanks for your comment.

  12. jad says:

    Today’s Times Colonist has riding maps for both EJDF and Saanich-Gulf Islands showing how the vote migrated.

    http://www.timescolonist.com/

    Quite clearly in both these ridings, the former third party benefited from the almost complete collapse of the Liberal vote. In EJDF for instance, Keith Martin’s former riding, the Liberal candidate did not win a single poll.

  13. Alison says:

    Thanks, Alice. I didn’t intend to request you do the math for me; rather your opinion on whether their method for determining whether vote splitting occurred has merit.

  14. Una says:

    Democracy eh? Not really something you can calculate. People do what they want to do. I can hardly wait for the financial metrics.

  15. We won’t get them for quite awhile, Una, so don’t wait up. ;-)

  16. This data isn’t worth much without including some of the obvious factors pundits mention all the time during and after elections:

    - incumbents running, or not

    - scandals / incumbent disgrace

    - reigning provincial party and its popularity (impacts on federal)

    - regional issues

    - rural vs. urban issues

    - party-wide disgraces, e.g. sponsorship scandal

    Since the actual question for all advocates of tactical voting or vote swapping is how accurate are the polls just prior to elections and what stats should be combined to predict ridings, perhaps some attention to these factors would be helpful. Certainly incumbent factors are pretty easy to model.

Leave a Reply