Friday, September 15, 2006

Polls and predictions

Just the other day, I said I wasn't going to make any more predicitions (except for moving seats towards close to call) until my final prediction on Sunday. However, the reason for that was because a lot of things are in flux and I was going to use polls to help me make final calls in close races.

As a lot of my readers have pointed out, polls are often wrong and that is true (largely because they are a snapshot of time taken at least a few days before election day and because they have a margin of error). So, though my method is based mostly on instinct, history and reader tips, I am using polls to push a result one way or the other if I can't quite make a decision by myself.

Today, Omnifacts Bristol released their final poll, something I was not expecting until tomorrow.
So, ladies and gentlemen, I am going to make a semi-final prediction. I will still leave some seats as leaners but I am throwing out my rules of moving a seat only one position per day and will play to leave these results as is (moving the leaners to safe) on Sunday. However, this will give me two days grace where my kind readers can warn me of blunders.

So, first of all, here is a poll round up:

Omnifacts Bristol (Sept 10-13, margin of error +/- 3.1%) - link

Decided voters: PC 30%, Lib 29%, NDP 5%, Other 1% (undecided 27%, refused 8%)
Decided voters + leaners: PC 46%, Lib 46%, NDP 7%, Other 1%

They also include very good regional breakdowns which I will elaborate upon later.

CTV Atlantic/Innovative Research Group (Sept 8-10, margin of error +/- 3.46%) - link (PDF)

Decided voters: Lib 32%, PC 31%, NDP 8%, Other 1% (undecided 14%, refused/not voting 15%)
Just decided voters: Lib 45%, PC 43%, NDP 11%, Other 1%

They provide less detailed regional breakdowns.

L'Acadie Nouvelle/Corporate Research Associates (Sept 7-10, margin n/a)

Just decided voters: Lib 44%, PC 42%, NDP 10%


My read of these numbers is a dead heat with a statistically insignificant edge to the Liberals. The Omnifacts Bristol poll, the only poll the surveys undecideds for how their vote is leaning, shows that the Liberals gain ground with leaners.

Considering that the Liberals lost the popular vote by 1% in 2003 and, with that, would have won under these ridings, then I think things are looking pretty good for the Liberals.

Here are Omnifacts Bristol's regional numbers:

Northwest (Victoria-Tobique, Grand Falls-Drummond-St. André, Restigouche-la-Vallée, Edmundston-St. Basile, Madawaska-les-Lacs, Campbellton-Restigouche Centre)

PC 53%, Lib 38%, NDP 9%

My analysis: This is a substantial lead for the PCs, though they are expected to be doing well here due to the large marjorities they will take in Madwaska-les-Lacs and Edmundston-St. Basile, I think they will definitely be claiming Restigouche-la-Vallée and possibly Campbellton as well.

Northeast (Dalhousie-Restigouche East, Nigadoo-Chaleur, Bathurst, Nepisiguit, Caraquet, Lamèque-Shippagan-Miscou, Centre-Péninsule, Tracadie-Sheila, Miramichi Bay)

Lib 55%, PC 38%, NDP 6%

My analysis: An even wider lead here than for the Grits here than the Tories have in the Northwest, this suggests that perhaps the Liberals can hold on to Centre-Péninsule and will hold Bathurst. The low percentage for the NDP means that either Caraquet was under sampled or they are not as much of a threat there as some suggest

Central (Carleton, Woodstock, York, York North, Fredericton-Nashwaaksis, Fredericton-Fort Nashwaak, Grand Lake-Gagetown, Oromocto, New Maryland-Sunbury West, Fredericton-Silverwood, Fredericton-Lincoln, Southwest Miramichi, Miramichi Centre, Miramichi-Bay du Vin)

PC 52%, Lib 38%, NDP 8%

My analysis: This large and bizarre conglomeration of ridings that are very heterogenous are hard to project based on these numbers. However, the large lead for the PCs will cause me to give them the edge in rural seats.

Southwest (Charlotte-Campobello, Charlotte-The Isles, Fundy-River Valley, Saint John Lancaster, Saint John Portland, Saint John Harbour, Saint John East, Saint John-Fundy, Rothesay, Quispamsis, Hampton-Kings, Kings East)

Lib 45%, PC 42%, NDP 10%

My analysis: I am surprised the NDP is so high here but they have a lot of strong candidates which may get 15 or so percent in a lot of ridings, I don't expect them to win anywhere.

Southeast (Petitcodiac, Albert, Riverview, Moncton North, Moncton West, Moncton East, Moncton Crescent, Dieppe Centre-Lewisville, Memramcook-Lakeville-Dieppe, Tantramar, Shediac-Cap-Pelé, Kent South, Kent, Rogersville Kouchibouguac)

PC 48%, Lib 46%, NDP 5%

My analysis: If you factor out ridings where the Liberals aren't players and ridings where the Tories aren't players, I think it gives a slight edge to the Liberals in battleground ridings

A post will follow with changes.

No comments: