Saturday, May 19, 2007

Election oracle, part 1














Image: John William Waterhouse - Consulting The Oracle (oil on canvas; 1884)

This week's Free Press/Probe election poll included an interesting addition: a voter migration matrix. This is a matrix of voters by the party they supported in one election with the party they supported (or plan to support) in a second election.

The poll data showed 2007 provincial election voting intention by reported 2003 votes cast. The poll showed for example that 64% of 2003 NDP voters plan on voting NDP again, while 9% will vote PC, 4% will vote Liberal, and 22% are undecided. The Tories appear to be keeping a greater share of their 2003 voters (69%), while the Liberals are struggling to hold their own, with just 44% of their 2003 voters intending to vote the same way this time. Twenty-nine percent of 2003 Liberals remain undecided, which is higher than for any of the other parties.

For fun, I recalculated the poll numbers to exclude the undecided and entered them into UBC's voter migration matrix election forecaster. Developed by Professor Werner Antweiler, the forecaster projects the voter migration numbers you enter into overall and seat-by-seat results. In response to my entry, the forecaster spit out an election prediction of 44.5% NDP, 39.4% PC, 12.8% Liberal, and 3.3% other, which isn't unreasonable.

The forecasted seat-by-seat results showed 35 NDP, 22 Conservatives, and zero Liberals. Fort Garry is predicted to shift from NDP to PC, while the Liberals are predicted to lose Inkster and River Heights, the former to the NDP and the latter to the PCs. Tight holds for the NDP were shown in St. Norbert, Seine River, and Radisson, while Minnedosa remained a tight hold for the PC. Other expected tight races generally weren't.

The calculator's weakness is that it applies voter migrations universally across all constituencies: in reality, voters may swing in greater numbers or to different parties in some regions than in others. It also can't account for the effect of "star" candidates, the added name recognition that comes with incumbency, the degree of effort put into winning by a party or candidate, or local issues. The numbers I entered also didn't take into account the weakness of the Liberals, who are holding less than half of their 2003 vote and whose intended voters this time are by far the softest of any party's.

For years, I've played with how shifts in party support province-wide translate into shifts at the constituency level. I'm always surprised at how the shifts are generally parallel despite so many local factors that must also play a role. For example, if a party increases their province-wide support by one-fifth (say, from 20% to 24%), it's likely that the same proportional increase will materialize in each constituency (say, from 5% to 6% and from 40% to 48%). An Excel-based calculator I created that does this automatically provides a good baseline for making seat-by-seat predictions. A consideration of incumbency, the level of effort a party is putting into a seat, and regional trends or local issues has to be made as well, but amazingly, these usually account for less than 5% of the vote, according to some studies that try to weigh the impact of these variables.

Right now, I'm predicting province-wide totals of about 45% NDP, 40% PC, and 14% Liberal, with relatively little change in seats between the parties. The NDP could lose a non-Winnipeg seat or two to the Tories, but then make up for it by picking up Inkster or Kirkfield Park. I'll profile the seats I expect to change hands or be close races over the next couple of days.

Strategically-speaking, the Tories simply must close the seat gap between the NDP and themselves (a difference of 15 in 2003) to come out of the election looking at all credible and able to challenge in 2011. If the Tories fail to pick up any seats outside the Perimeter Highway from the NDP, who then pick up one or two inside the city, it'll be completely devastating for Manitoba's party of the right. Tories tend to be a lot harder on their leaders than do New Democrats, so I think Hugh will be in for a very rough ride if he doesn't pull off victories in at least a couple of seats they don't already hold.

Those are all my thoughts for a late Friday evening. There'll almost certainly be more to say tomorrow.

No comments: