Monday, May 09, 2005

Electionprediction.org and Seat Security

Along with nodice.ca and politicswatch, electionprediction.org is a must-view for those obsessed, as I am, with the potential for a spring election. At EP.org they are fairly good predictors--see the track records posted on the front page of the site--and are an excellent source for information on the "local colour" that can determine close races or point to races that are more about the local candidates than about the national policy, polls, and leaders.

They have now begun accepting comments and making predictions for the yet to be called 2005 federal election. The choices for predictions are Liberal, Conservative, NDP, Bloc. or too close to call (TCTC). After a casual inspection I've noticed that they are understandably conservative (in the very small-c, non-political sense of the word) in their predictions. As far as I can tell the choices are either that the incumbent holds the seat, the race is TCTC, or there isn't enough information (comments) to make a prediction. I think it would be interesting to see which party's seats are considered the most secure, and therefore which other parties have the most opportunity to improve.

Here are the results of my data collection:
NationalIncumbentTCTCNo CallTotal
LIB724915136
CPC77212100
NDP134219
BQ3911353

Notes:
David Kilgour counted as a CPC. I know he's not but they have already called this as a CPC seat so if I were to count it otherwise it would be the only predicted incumbent upset.
Chuck Cadman counted as a CPC. His riding is right now called as TCTC, but the only race would seem to be between Cadman, if he chooses to run as an independent and whomever the CPC fields.
Carolyn Parrish counted as a Liberal.
Even though it's vacant Labrador counted as LIB.


So, by dividing the number of Too Close to Call seats held by each party by the number of seats that the party currently holds we can see that the raw vulnerability index for the parties is:

PartyVulnerability Index (% of current seats)
Liberal47.06
Conservative23
New Democratic Party31.58
Bloc Quebecois26.42

The problem here is that No Calls are counted the same as races that are too close to call (i.e. against the incumbent). It seems that there are a few reasons for a "No Call" either a riding is very predictably safe for the incumbent and therefore no partisans have an interest in either "pumping" or "dumping", or a riding has a possibly retiring incumbent or the candidates haven't been picked. Also, there seem to be a high number of No Calls in Ontario and Quebec. The Ontario number may be because there are just so many ridings. If someone considers themselves an expert on Nova Scotia politics it is easy for them to post a comment on all 11 Nova Scotia ridings and thereby trigger a prediction on all of them--EP.org seems unwilling to make a prediction until a comment is made so it may just be that the Ontario "experts" haven't gotten around to commenting on all of the ridings. The same numbers argument can also be applied to Quebec. Additionally, because most of the Quebec No Calls are concentrated in BQ ridings we might assume that a language barrier exists--even though many comments are in French.

I think it would be more accurate to produce an Adjusted Vulnerability Index by counting the No Calls in favour of the incumbent. This produces:

PartyAdjusted Vulnerability Index (% of current seats)
Liberal36.03
Conservative21
New Democratic Party21.05
Bloc Quebecois1.89

No big surprises here. The Liberals and BQ reap the most benefits from this adjustment. We know that the Liberals are strong in Ontario, especially in urban and suburban ridings that are more likely to swing. Also, no one was expecting the Liberals to win many (or any?) seats from the BQ so no one should be shocked to see the BQ number so low. I'll be posting my conclusions and interpretations in another post this afternoon.

0 Comments:

Post a Comment

<< Home