The latest update to The Writ’s Vote and Seat Projection continues to show a big advantage for the Liberals, with some of last week’s slippage reverting back to the party.
It suggests that the holding pattern in voting intentions that we’ve seen for the last two months is continuing unabated.
The Liberals lead in the projection with 46.2% of the vote, up 0.5 points from last week. If we round to the nearest full percentage, the Liberals have been holding at 46% in the projection since the March 3 update. The Conservatives are also up 0.5 points to 34.5%, while the New Democrats have dropped 0.6 points to 8.8%.
The Liberals are leading in 221 ridings in the projection. Taking into account potential polling and modelling errors, the Liberals are projected to win an average of 211.4 seats. Compared to last week, the Liberals are leading in three more seats but their average number of projected wins is up 1.4 — a reflection of how the Liberals have moved ahead in several toss-up seats that might not go their way in an actual election.
The Conservatives are leading in 86 seats, unchanged from last week, and win an average of 103.4, up only marginally. The NDP has lost the lead in three seats in the projection and is down 1.2 seats on average to just 7.9. The Bloc Québécois is holding steady with 19.4 projected seat wins (while they lead in 22).
For fully interactive charts, tables and maps, head to The Writ’s website:
I’ve had some questions about the difference between the “Seat Projection” and the “Avg. Projection”. Let me try to explain.
The model swings the results of the last election in each individual riding according to where the polls are today. Riding-level factors — such as the presence of an incumbent — are then taken into account. The Seat Projection shows the number of ridings in which each party is leading according to these calculations. This is, in other words, an estimation of how many seats each party would win if both the polls and the modelling was 100% accurate.
Of course, we know that 100% accuracy is impossible. Even if the polls were entirely correct, the model will make some riding-level errors because it is based on the assumption that all ridings will behave in a similar fashion. But some ridings will behave slightly differently than others — a party’s gain in Quebec might be more dramatic in some regions than in others even if, on average, it all comes out in the wash.
Then there is the error that we should expect in the polls. The polls will always be off by at least a little bit, but we don’t know in which direction. Will they over-estimate the Liberals and/or under-estimate the Conservatives, or vice-versa? The Seat Projection is agnostic on this, assuming that it is equally likely that the polls will be off in one direction or another for each party.
The Avg. Projection, however, is not agnostic about this. Using the past five elections as a guide, it assumes, for instance, that the Conservatives will beat their polls most of the time. For that reason, the Conservatives will be considered in contention to win more seats in which they trail than will the other parties. The Avg. Projection looks at all those seats where multiple parties are in contention, and assigns them randomly to one party or another in 10,000 simulations.
So, let’s apply this to this week’s projection. The Liberals lead in 221 seats and the Conservatives lead in 86 seats. But the Liberals are leading in more closely-contested seats than are the Conservatives. And, with the Conservative tendency to beat their polls, it seems likely that some of those toss-ups in which the Liberals lead will fall the Conservative way. Similarly, it is more likely that the Conservatives will hold more of the toss-ups in which they lead than would the Liberals.
As a result, the Liberals are projected to win an average of 211.4 seats, with the Conservatives winning 103.4 seats. In other words, the Liberals are expected to lose more of the closely-contested seats in which they are in contention than they win, while the Conservatives are expected to win more of those closely-contested seats than lose.
The model doesn’t “know” which seats these are, as this process is abstracted. But you can see which seats are the most likely to go either way in the Riding Projections. The “Toss” seats are the most uncertain — and it’s worthwhile to search them out. The Liberals, for instance, are projected to be leading in nine seats in Alberta, which would be quite a haul. But four of them are classified “Toss”. Based on historical precedent, that means they could be tough to actually win in practice, especially considering the Conservatives’ tendency to over-perform their polls by significant margins in Alberta. Two other Liberal seats in Alberta are “Edge” seats, the next-most uncertain classification.
So, even if the Liberals are projected to be leading in nine seats in Alberta, the real-world of an election might deliver them far fewer than that. This is what the Avg. Projection reflects.
Turns out, elections can be unpredictable, messy things, even in the best of times!