À propos

DemocraticSPACE is Canada’s leading source for non-partisan election news coverage. Offering our unique seat projections, riding-by-riding discussion forums, poll tracking, and election analysis, DemocraticSPACE is literally where democracy happens^TM. Featured in national media across the country, including the CBC, Toronto Star, Montreal Gazette, Calgary Herald, among many others, our audience is drawn from across the political spectrum and includes pundits, reporters, campaign staffers, elected officials, and of course, Canadians interested in objective coverage of local, provincial, and federal elections. If it’s about elections, DemocraticSPACE is where it happens.

DemocraticSPACE Research (DSR) is our non-partisan research and consulting arm. Put the power of DemocraticSPACE to work for your campaign. Using our unique local and regional statistical modeling, we are your one-stop-shop for campaign strategy. We partner with local and regional pollsters to obtain data that is modeled at the regional and local levels, giving your campaign the most up-to-date, detailed spatial analysis available. Our modeling allows us to offer your campaign detailed recommendations from communication strategies to boots on the ground. Call or email us to discuss your needs.

DemocraticSPACE is lead by Gregory D. (”Greg”) Morrow. Formerly on the faculty in the Department of Urban Studies & Planning at the Massachusetts Institute of Technology (MIT), Greg is currently completing his PhD in the School of Public Affairs at the University of California, Los Angeles (UCLA). Greg is originally from Eastern Ontario, and educated in Québec. Greg’s research operates at the intersection of politics and urban space.

About Our Projection Model
We have constructed an intricate mathematical model to translate polling data into seat projections (since, due to our first-past-the-post electoral system there is often little relationship between popular vote and the number of seats a party earns in parliament). Thus, seat projections are the most reliable way to understand how we expect to win the election and by how much.

We make vote projections for every riding in Québec. These projections are based on province-wide and regional polling data by major Canadian pollsters, among others, Nanos, Ipsos-Reid, Angus Reid, Harris/Decima, Strategic Counsel, Ekos, Environics, Segma, Léger Marketing, Unimarketing, and CROP.

Individual polls have small sample sizes, leading to high margins of error, which leads to seemingly incongruous poll results from day to day. To provide more reliability and stability, we aggregate the surveys of all the pollsters over a narrow time period to provide a larger sample size, thus lower margin of error, to produce a weighted average of what the current polls are telling us (our aggregated data is weighted by sample size, so larger samples are weighted more heavily than smaller samples). To ensure the most reliability, it is important to balance larger sample size with a small time period. As such our weighted aggregated averages typically cover several days during the campaign.

Adjusting for Regions
We divide Québec in 13 sub-regions. We know that change in support is not uniform across all sub-regions. We account for this in our projections by looking at past results — comparing the movement of a given sub-region with the overall movement in the polling region. For example, we compare the movement for each party in each sub-region if it reflected the average of the polling region, then compare it to the actual movement. Inevitably, some regions swing more or less for one party or another. We apply factors to each sub-region, which are multiplied by the overall change in support with in a given region. For example, if the Liberals dropped 10 points in Québec, but only dropped 5 points in Montréal, we might expect any downward movement for the Liberals to be felt at only half the strength in Montréal. Likewise for every party in every sub-region. Recognizing that past performance is not a perfect predictor of future performance, some judgment is necessary. Moreover, when sub-regional and riding polls are available, we adjust our sub-region factors to reflect what the polls are telling us. However, in most cases, sub-regional polling data is unavailable, hence the need for mathematical modeling.

The model is further refined by dividing ridings within each sub-region into average, above average, and below average for each party, since not all ridings within a given sub-region are the same. For stats fans, the average ridings are those in which the change in support within plus or minus one standard deviation from the average in the sub-region. Above average are those that are above the average by more than one standard deviation and vice versa for below average ridings. These are typically small adjustments.

Finally, the model adjusts for candidates and intangibles for local hot-button issues. This requires taking into account past star candidates who are not re-running, new star candidates, incumbents who are not re-running, among other intangibles. Sometimes, local hot-button issues can swing a particular riding one way or the other. We make our best guess at how these intangibles might impact the race.

The reliability of primarily using party preference as the backbone of the mathematical model is due to the fact that the vast majority of Canadians cast their ballots according to party preference, rather than local candidates. According to past surveys, on average, about 80-85% of Canadians cast a party vote (60-65% on party and another 15-20% based on the party leader), while only 15-20% of voters decide how to cast their ballot based on the local candidate.