A record of forecasts vs. results can be found below - the barometer has yet to make a wrong call in 29 federal and provincial by-elections.
What the By-Election Barometer is
The By-Election Barometer tracks all scheduled and upcoming federal and provincial by-elections. The percentages shown in the charts represent margins, colour-coded according to the standards adopted by this site. The first set show the results of the last two elections.
The second set of margins are those that the projection model churns out when each regional/provincial poll is applied, using the same system as ThreeHundredEight's standard seat projection model. These are a way to demonstrate what might be expected in the riding, based on regional trends. The rolling 30-day average calculates an unweighted average of the last 30-days of regional polling.
The third set of margins, when available, represent the results of actual polls of the riding that have been released.
Finally, ThreeHundredEight's Forecast is calculated by taking the projected range of results from the average of the last 30 days of polling, including any polls done for the riding itself, and comparing how the ranges for each party overlap. The amount of overlap that potentially puts a party in a position to win is then tallied, the result being a percentage "chance" of that party winning the riding.
A Strong result means a 95% to 100% chance of winning, Likely is a 75% to 94% chance, Lean is a 60% to 74% chance, and Toss-Up means the chances of a party winning are 59% or less.
The Wildcard section describes any factor that could make the result unpredictable.
What it isn't
The By-Election Barometer is not a poll, the section titled "Margin after application of swing from regional polls" is not a list of riding polls, and the 30-day average is not a projection. By-Elections are notoriously hard to call, and the Barometer is not a tracking of actual voting intentions. The forecast is also not an opinion. As always, I am tied to what the numbers show.
By-elections have yet to be called in the federal ridings of Macleod and Fort McMurray-Athabasca in Alberta and the provincial riding of Virginia Waters in Newfoundland and Labrador.
Weighted averages and support ranges
When available, polls for by-elections have been weighted using the same aggregation methodology employed elsewhere on the site. The confidence intervals are based on the standard deviation from the mean of past errors in by-election polling. What this means is that based on how by-election polls have been wrong in the past, we can expect the results to fall within the 68% confidence interval (one standard deviation) 68% of the time, and within the 95% confidence interval (two standard deviations) 95% of the time.
This is in contrast to the margins of error normally reported in by-election polls: often +/- 4%, 19 times out of 20. While this may be the correct margin of error to apply to a true random sample, recent experience has shown that this confidence interval has not been reliable. Instead, to have confidence that the results will be accurate 19 times out of 20, we need a margin of error of about +/- 8%. The confidence intervals calculated below reflect this.
It is hard to find a safer riding for the Conservatives than Macleod. Menzies won it with 77% of the vote in both the 2008 and 2011 elections, after taking it with 75% in 2004 and 2006. Between the 1988 and 2000 elections, the combined tally of the Reform/Canadian Alliance and Progressive Conservatives in the riding ranged between 81% and 84%. The best the Liberals have done since then was just 16% in 1993, while the NDP has not done better than the 10% it took in 2011 over that time.
Macleod is as secure as it gets. The only battle in this riding will be for the Conservative nomination. Who places second may be an interesting footnote, but that is about it.
Forecast history: Has been, and will remain, STRONG CONSERVATIVE from Menzies' resignation.
The Conservative Party and its predecessors have held the riding since the 1950s. In 2011, Jean took 72% of the vote to just 13% for the NDP and 11% for the Liberals. The last time the riding was not won by a majority of the vote was in 1993, when the Reform candidate captured 47% of it.
So it is unlikely that either the Liberals or the New Democrats will be able to make a major run at the riding. But after the results in Provencher and Brandon-Souris last year, we should not under-estimate the ability of the Liberals to make some serious inroads.
In fact, recent polling in Alberta has been quite favourable for the Liberals, and the swing applied from two recent surveys would put the Liberals in range of being able to win. That the rolling average gives the Tories the victory by a little more than 10 points, after winning it by more than 50 in the last few elections, is remarkable itself. The riding is still heavily leaning Conservative, but it has been re-classified as a Likely Conservative win.
Forecast history: Was STRONG CONSERVATIVE from Jean's resignation. LIKELY CONSERVATIVE on February 21.
The results in Virginia Waters will be interesting to see, because the riding has been a bellwether since at least 1975 (if we stretch its history back into the riding of Pleasantville before 1996). Since then, its MHA has always been on the government side of the House of Assembly.
The riding has voted reliably with the government (or the party about to form government) since it was created in 1995. The Liberals took between 56% and 47% of the vote in their wins in 1996 and 1999, while Dunderdale of the Tories took 58%, 73%, and 60% in the elections of 2003, 2007, and 2011. The New Democrats finished second in the riding in the last two elections, with 13% in 2007 and 30% in 2011. The Liberal vote stood at just 10% in that last vote.
Where will the Liberals sit when this by-election is held? The latest poll from CRA gives the Liberals a very wide lead in the province. But from the model's point of view, the party is starting from a very low base in Virginia Waters. This gives the PCs the potential to hold it, but the Liberals are not ruled out. And with the party surging in the polls, they stand the best shot at landing a good candidate and wrestling it away from the Tories. If the NDP's vote collapses and swings to the Liberals, they might very well win it. Nevertheless, the model gives the Tories a slight edge.
Forecast history: Was LEAN P.C. at Dunderdale's resignation.