Projecting the 2021 Election, One Week Out

(Short on time? The projection is at the end of the post – search for “one-week projection” or scroll down.)

Over the past few elections, the increasing number of polls being published and the increasing sophistication of polling aggregators like CBC Poll Tracker and 338Canada allow us to enjoy a much more accurate read on the intentions of Canadian voters.

But wouldn’t it be great if we could go a step further and predict voting intentions ahead of time? What if we could reliably project the outcome of an election, say, a week before voting day?

The final week of a campaign is often an anticlimactic one. Debates are finished, platforms have been published and dissected, and voters are getting tired of it all. And yet, we often see significant changes in support on voting day, with parties mysteriously jumping or dropping as much as several percentage points.

Why is this? Is it simply that polls are inaccurate? Or are there perhaps patterns at work—patterns that we could analyze and predict?

Election day gains and losses, 2000-2019

The graph below plots the results from the last two decades of Canadian elections. For each party in each election, it compares the party’s popular support in the final few polls of the campaign (x-axis), versus the party’s actual election result (y-axis). Tap or click for full size.

Graph comparing a party's popular support in the final polls of a campaign to its actual election result, for all parties polled for elecitons in the year 2000-2019. Key elements of the graph are described in subsequent paragraphs.

Immediately we can see that there is a definite shape to this data. Parties with higher popular support in the closing days of the election are more likely to outperform their polling, while parties with lower popular support are more likely to underperform their polling.

To see this effect in numbers, we can divide the data into two buckets. The first bucket will contain every instance of a party that had less than 25% popular support going into election day, and the second bucket will contain every instance of a party that had more than 25% support. This gives us the following averages:

Popular support in final polls of campaign<25%>25%
Average gain or drop on election day-0.7%+1.0%

This dispels the idea that gains and losses are due to “noise” in polling. Noise would have no shape at all. So what’s causing this effect?

Is it “strategic voting”?

The term “strategic voting” has several meanings. Here I’m using it to describe a behaviour where a voter switches their support away from their preferred option to a less-preferred option, in order to “block” a win by an even-worse third party.

For example, in the 2021 election, there has been pressure on NDP and Green supporters to vote Liberal to “block” a Conservative win, and pressure on People’s Party supporters to vote Conservative to “block” a Liberal win.

Is this type of “strategic voting” causing larger parties to gain support on election day, and smaller parties to lose support?

The following graph shows the same data as above. This time, each point has been categorized according to whether the election was a close race or a blowout. (I’ve defined a “blowout” as more than 5% separating the top two parties; less than 5% separation means a “close race”.) Tap or click to view full size.

Graph once again comparing a party's popular support in the final polls of a campaign to its election day results. Data points are visually grouped according to whether the top two parties were separated by more or less than 5% in the final polls. Data includes all polled parties in elections in the years 2000-2019. Key elements of the graph are described in subsequent paragraphs.

On first inspection, there isn’t any striking difference between these two groups of data. The trend line for the close races is a hair steeper, but there are individual data points all over the place.

Let’s divide these results into four buckets: parties with less than 25% support in a close race, more than 25% support in a close race, and then the same support levels in a blowout. That gives us the following averages:

Popular support in final polls of campaign<25%>25%
Average gain or drop (same as previous table)-0.7%+1.0%
Average gain or drop in a close race-0.8%+1.3%
Average gain or drop in a blowout-0.6%+0.7%

Again we see a slight difference, but it’s far from explaining the full magnitude of election day gains and losses. In particular, the support lost by small parties is barely affected by the closeness of the race—something which goes directly against what we’d expect if “strategic voting” was the root cause.

We’ll need to look further to explain the gains and losses that parties experience on voting day.

Is it momentum?

Often times you’ll hear political pundits refer to momentum, the idea that success breeds more success, and failure breeds more failre. But does momentum really exist? Or is this a bad sports metaphor that doesn’t apply?

The graph below plots the same set of election results from 2000-2019, this time comparing each party’s change in support during the last week of polling to its gain or drop on election day. Tap or click to view full size.

Graph comparing a party's change in support during the last week of polling, versus the party's change in support from final polls to election results. Data is from elections in the years 2000-2019 for all polled parties. The data shows no particular pattern.

Clearly, there is no pattern here. A party can spend the final week steadily gaining or losing support, and then have the opposite thing happen on election day. Momentum does not appear to exist—at least, not in the final week of the campaign.

It’s about performance

If election day gains and drops can’t be explained by polling inaccuracy, “strategic voting”, or momentum, then what’s the cause?

There’s a simple answer: performance.

Let’s revisit our data again. This time, we’ll visually group results according to political party. Tap or click to view full size.

Graph once again comparing a party's popular support in the final polls of an election with its actual election results. Data points are grouped visually according to party and show a distinct pattern of gains and drops for particular parties. Data is for all parties polled for elections in the years 2000-2019. Key elements of the graph are described in subsequent paragraphs.

There are clear patterns here. The two larger parties rarely drop by much on election day. Meanwhile, the NDP and Greens always drop. (In the case of the NDP this rule holds whether they’re polling at 8% or 33%.) Meanwhile, the BQ, despite having less national support than the NDP, has a mixed result most closely resembling the Conservative party.

What causes these distinct clusters? Is it the unique character of each party, or is there an overall explanation that applies to all of them?

One possibility is that larger, richer parties are able to run more consistent “get-out-the-vote” efforts on election day. (GOTV is a system of co-ordinated, mass phone calls and texts to previously-identified supporters, reminding and encouraging them to vote.)

Stronger GOTV would result in a greater number of a party’s supporters actually voting, while weaker GOTV results in a smaller number of supporters voting. In the final election results, these differences would look like “gains” and “losses”.

What about the Bloc?

But what about the BQ? Where do they fit into this theory, given that they’re a smaller party, but they often avoid losing support on election day?

Remember that the BQ only runs candidates in Québec. As a result, whatever resources they spend on GOTV are spent only in Québec’s 78 ridings, rather than being spread across the full 338 ridings contested by the other parties.

A quick comparison of the resources available to each party shows that, despite only receiving the fifth-highest total contributions, the BQ is the third-richest in contributions per riding they contest.

PartyContributions (2020)Ridings$/Riding

When we sort the parties this way, we see that the BQ can be thought of like a large party operating in a limited geographical area. Its concentrated GOTV would explain the gains and losses that more closely resemble the Liberals or Conservatives, rather than the NDP and Greens.

What if we return to our original graph, but remove the BQ data points? (The faint curve represents the data with BQ included, the dashed curve with BQ excluded.) Tap or click to view full size.

Graph once again comparing a party's popular support in the final polls of an election with its actual election results. Data is for all polled parties in elections from 2000-2019, with the exception of the Bloc Québécois, which has been removed. The data now shows a more consistent shape, with larger parties gaining and smaller parties dropping on election day. Key elements of the graph are described in subsequent paragraphs.

With the BQ removed, the shape of the data fits our theory of GOTV performance nicely. (We might even imagine the BQ’s data points being reinserted around the 20% mark, just right of the NDP and somewhat left of the Liberals and Conservatives, as fits their per-riding resources.)

We can also revisit our graph of close races versus blowouts with the BQ removed. Tap or click to view full size.

Graph once again comparing a party's popular support in the final polls of an election versus its actual election results, for all parties polled in elections from 2000-2019, with the exception of the Bloc Québécois, which has been removed. Data points are visually grouped according to whether the top two parties in the election were separated by more or less than 5%. Key elements of the graph are discussed in subsequent paragraphs.

(Note that this graph is flawed. We’ve removed the BQ, but we haven’t compensated by removing the Québec ridings for the other parties, nor have we compensated for the fact that the BQ is the leading party in many Québec ridings and could benefit from “strategic voting”. Those calculations are too complex for this post, but could be explored in a future model that simulates Québec separately from the rest of Canada.)

The situation is much the same as before. The difference between close races and blowouts is a bit more pronounced, but “strategic voting” is still clearly not the dominant factor. In close races, it might account for up to a third of gains and losses on election day.

You might also notice that close races appear to be more volatile, with greater gains and losses, regardless of support level. This makes sense, because in a close race, voters might be on the fence between the two leading parties and break one way or the other in large numbers.

Possible Models for the 2021 Election

So, what will happen this year? We’re 7 days out from election day. Can we take today’s polling data, and feed it into our model to estimate a final result?

We have a few choices for how we might project from here to election day:

  • We could use each party’s historical average for election-day gains or losses.
  • We could project based on a generic curve that relates support to election-day gains and losses.
  • We could observe that this election is a very close race (with only a 0.6% gap between front-runners), and use a curve adjusted for close races, to take into account a modest “strategic voting” effect.

Why not do all three? Below are the results of each projection. (Current support levels are an aggregate of polls, drawn from today’s update of the CBC Poll Tracker.)

Current support (7 days out)31.9%31.3%19.3%6.5%6.4%3.6%
Projection, based on …
… historical performance33.6%32.9%18.0%5.6%6.5%2.5%
… curve based on party size33.4%32.7%19.1%5.6%5.5%2.7%
… curve adjusted for close races33.9%33.3%19.7%5.2%5.1%1.9%

The three models are largely in agreement, with some variation in the NDP, BQ, and Green totals.

2021 Election: One-Week Projection

To take this a step further, we might try a blended model that combines historical performance and the “close race” curve, but omits the curve for the BQ because of their unique situation.

Using a seat-count simulator, such as the interactive simulation on, we can translate our projection into seat totals. Using the blended model, we get:

Current support (7 days out)31.9%31.3%19.3%6.5%6.4%3.6%
Projected gain or loss+1.6%+1.5%-0.6%-1.2%+0.1%-1.4%
Projected result33.5%32.8%18.7%5.3%6.5%2.2%
Projected seats (338Canada)165125280191

Under this projection, the Liberals gain 10 seats, but fail to achieve the majority they hoped for, with an overall performance very similar to the 2019 election. Among the other parties, there are winners and losers:

  • The Conservatives gain 6 seats, but lose the “moral victory” they had achieved in 2019 when they had the highest popular support of all parties.
  • The NDP gain 4 seats and build significantly on their popular support from 2019. While not a breakthrough, the outcome is a solid gain.
  • The BQ suffer a significant disappointment, dropping 13 seats and about an eighth of their popular support.
  • The PPC establish themselves as a minor party, but fail to win seats due to their spread-out base of support. If they can maintain support levels, inclusion in the next elections’ debates is likely.
  • The result for the Greens is devastating, with their caucus reduced to one (former leader Elizabeth May), and their lowest popular support in twenty years. Even worse, their support is below their “tragic number” of 2.4% and they fail to secure up to $1M of public funding—the only party to win seats but not qualify.

Check back after the election, when I’ll compare the actual results to these projections. Will the model hold up? And if it des, can we extend it to a general model that performs well on all elections from 2000-2021?

Greens Fail to Nominate Full Slate; up to $1,000,000 of Public Funding At Risk

The deadline to nominate candidates for the September 20th election has passed, and for the first time in two decades, the Green Party has failed to nominate a full slate.

On August 15th, I raised the possibility that the Greens might fall short, nominating somewhere in the neighborhood of 245 candidates. As I write this, two days after the deadline, Elections Canada lists (by my own count) 248 confirmed candidates for the Greens, leaving 90 ridings uncontested.

Certain regions are particularly badly hit: Newfoundland and Labrador have no Green candidates whatsoever; Québec is missing candidates in one third of its ridings; Alberta is missing half. Toronto and surrounding area are riddled with holes, especially puzzling given their proximity to leader Annamie Paul’s riding.

The party’s slate may include as many as 100-120 “paper” candidates (people who have volunteered to put their names on the ballot, but do not intend to actively campaign). The last time the party reported a candidate total to the media, prior to an emergency recruitment drive for paper candidates in the closing week of nominations, the total sat at 142 candidates. (As of today, only 145 candidates are displayed on the party’s website.)

What Happened?

A Global News report of September 1 indicates that the party had prepared a strategy in December to put forward a full slate, and, according to leader Annamie Paul, “enough candidates [stepped] forward”. However, “lack of resources” were blamed for the party’s failure to collect the 100 signatures per candidate required to run.

Paul also stated that the party “took a little more time getting off the ground … in order to make sure that we were more diverse and inclusive”. (Several party officials have privately described the candidate search to me in a less complimentary manner, calling it “slow” and “bureacratic”.)

All signs indicate that the party fell into the “slate canyon”, a phenomenon I described in my previous post. Parties in the 2-5% range of support must either commit to running a full slate, or a smaller slate of their strongest candidates, or they risk serious financial consequences.

A Million-Dollar Mistake?

A major source of public funding for federal parties in Canada is the campaign expense reimbursement. To qualify, a party must meet one of two criteria:

  1. They received 2% of all the votes cast in the national election.
  2. They received an average of 5% of the votes cast in the ridings where they ran candidates.

Parties that qualify receive a reimbursement for 50% of their central campaign expenses. In the case of the Greens, that could mean up to a million dollars.

In the 2015 election, the party’s expenses totalled $3.91M (however, this was a historically long election), and in 2019, they spent $2.45M.

In 2021, the party has experienced significant financial troubles, and earlier this year members discovered the party was mere months from going bankrupt, with only $300,000 cash on hand. However, despite those troubles, the party was able to secure a campaign loan of “under $2 million”, according to an August 22nd report in the Toronto Star.

To estimate the party’s likely campaign expenses this year, we can exclude the $300,000 cash (needed for survival expenses). Donations typically spike during an election, so combined with the campaign loan, we might expect an upper estimate of $2M for the party’s election expenses in 2021.

(A recent mailout by the party refers to a “campaign [fundraising] goal of $600,000”. It is impossible to know whether this goal will be reached, but it seems unlikely it will be exceeded. This gives further support to the upper estimate of $2M.)

That would make the 50% reimbursement worth $1M, or about half the value of typical donations in a non-election year.

How likely are the Greens to receive the expense reimbursement? This is where the reduced candidate slate becomes critical.

Calculating the Greens’ Tragic Number

Achieving the 5% threshold is unrealistic for the party at this point. With the party currently polling at 3.4%, with all the internal turmoil and financial strain, and with an established pattern of consistently losing support during election periods, there is no reason to expect the party to rebound to 5% overall support.

So, to secure the funding, the only chance for the Greens is to meet the other threshold by receiving 2% of all votes cast in the election. This is where the reduced candidate slate can be fatal, because it denies the party some of the votes it would have received with a full slate.

So if more than 2% support is needed because of lost votes, what would be the “tragic number” for the Greens?

The “Equal Ridings” Method

As a first guess, we can assume that all ridings are equal. Given that there are 338 ridings total, this means that any riding without a candidate would result in 1/338th of total votes being lost.

With only 248 candidates out of 338 ridings, the Greens would be losing 27% of their potential votes. To achieve 2% of the total votes cast, they would actually need 2.73% support across the country. The 0.73% would be lost in ridings without a candidate, and the remaining ridings would be enough to achieve the 2%.

“Worst Ridings” Method

However, all ridings are not equal, nor even close. So, for a more refined approach, we might assume that the 90 ridings without candidates are the party’s 90 weakest ridings. (This makes sense, since candidates are typically harder to find in ridings with less support.) The party would lose fewer voters from these ridings than it would from the average riding.

The following graph shows the Green Party’s vote concentration in the 2019 election. (Tap or click to view full size.)

Graph showing the distribution of the Green Party's vote in the 2019 election, with the 90 weakest ridings highlighted. Key details follow.
Vote concentration for the Green Party of Canada in the 2019 election.
Ridings are arranged in order from highest to lowest vote percentage received. The 90 weakest ridings are highlighted.

In 2019, the Green Party’s 90 weakest ridings brought in only 10.4% of the party’s total votes. (Much less than 27% of the total votes, as we would have expected if all ridings were equal.) So, to receive 2% of votes cast this election, we would estimate the Greens would need 2.23% actual support. The 0.23% would be lost in ridings without a candidate.

A Best Estimate

The “weakest ridings” model isn’t completely realistic either. For example, we know that in 2021, the party has no candidates in Newfoundland and Labrador. But in 2019, three of NL’s seven candidates receive more votes than the candidates in the party’s 90 weakest ridings.

We could refine our model by going riding-by-riding through the 2019 results, or by looking at regional polling today, but I don’t think that level of detail is necessary. It’s safe to say that the party’s tragic number lies somewhere between the “all ridings are equal” estimate and the “90 weakest ridings” estimate&mdash;likely closer to the latter.

My best estimate for the tragic number for the Greens this election is 2.4% support. Any less than this, and they would miss both threshold for the expense reimbursement.


The Green Party’s failure to nominate a full slate of candidates puts up to $1,000,000 of public funding at risk. To ensure that it receives the campaign expense reimbursement, the party will need to remain above 2.4% support by election day.

In past elections, the Green have consistently lost several points of support during the campaign. With the party in turmoil and under financial strain, there is every reason to believe the pattern will repeat this year.

With up to a million dollars on the line, the distance between 3.4% and 2.4% has never looked so small.

Updates & Sources

Updated September 3rd to add graph of the 90 weakest ridings, add the party’s campaign fundraising goal of $600,000, and make minor improvements to wording.

The Toronto Star reported the party had obtained a loan of “under $2 million”:

Global News reporting on the party’s nominated shortfall and some of the difficulties the party faced:

Campaign expenses in 2015 ($3.91M) and 2019 ($2.45M) are drawn from Elections Canada published returns.

Nominated candidates are listed on the Elections Canada website (by my count 248, as of the time of writing):

An easy-to-read list of candidates, organized by region, is available on Wikipedia (it showed 252 candidates at the time of writing):

The CBC Poll tracker aggregates polls from various polling companies to provide a more reliable measure of each party’s standings:

The party’s $600,000 fundraising goal was stated in a paper mailout to members, dated August 27, 2021. (Viewed first-hand.)

Slate Canyon: Expense Reimbursements and the Dilemma for Small Parties

In my previous post, I discussed the possibility that the Green Party of Canada may not run a full slate in the 2021 election, and could potentially lose hundreds of thousands of dollars of public funding as a result.

Why? The reason lies in Canada’s system of reimbursing campaign expenses from public funds. This post describes how those reimbursements work, and examines how the rules create a dilemma for smaller parties at a certain level of public support, while providing other parties with effectively guaranteed public funding.

Two Thresholds for Reimbursement

The Canada Elections Act provides a registered party with a 50% reimbursement on its election campaign expenses, provided the party achieves one of two thresholds:

  1. Get 2% of the votes cast across the country.
  2. Get 5% of the votes cast in those ridings where the party ran candidates.

Meeting either threshold qualifies the party to receive the expense reimbursement. It doesn’t matter how many candidates the party ran, or which ridings it ran them in, or how the votes were distributed. As long as the party’s support meets the threshold in the appropriate ridings, the party qualifies.

The Dilemma for Small Parties

For a party polling in the low single digits, these thresholds create a strategic dilemma:

  • To meet criteria #1, you should run the largest slate of candidates you can. Each additional candidate, even a very weak candidate, gives you more votes out of the total across all ridings, and gets you closer to the 2% threshold.
  • However, to meet criteria #2, you should run a slate of candidates whose average vote will be at least 5%. If your party is polling in the low single digits, this set of ridings would be less than a full slate&mdash;possibly much less.

The risk for a small party comes if it runs a mid-sized slate and its result falls somewhere between these two goals, achieving neither of them. To avoid this outcome, parties polling in the low single digits need to commit to meeting one of the two criteria and base their nomination and campaign strategies around that choice: they must cling to one of two cliffs to avoid falling into the “slate canyon” between.

An Example

The graph below illustrates the dilemma. The data for this graph is based on the real-world vote distribution of a party in the 2019 federal election, but scaled down so that its overall public support is equal to 2.1%.

Graph showing the vote distribution of a fictional party with 2.1% national support. Key elements of the graph are described in the following paragraph.
The “slate canyon” effect for a party with 2.1% public support and typical vote distribution for a small party.

From left to right, the party’s candidates are arranged in descending order of the vote share they’re expected to receive in their ridings. (The curve seen here is typical of smaller parties.) To read this graph, assume that whatever size candidate slate a party runs, it always picks the strongest possible candidates, so the slate expands from left to right across the graph.

For a particular size of candidate slate, the dotted line represents the percentage of votes received across all ridings. As the slate grows, this number increases due to the addition of more and more votes, reaching the 2% threshold when the slate includes the party’s best 292 candidates (out of a possible 338). Slates that achieve the 2% threshold are coloured blue.

Meanwhile, the dashed line represents the average level of support in those ridings where candidates were run. As the slate grows, this number decreases, due to the addition of weaker candidates. Average support falls below the 5% threshold for any slate larger than the party’s 37 best candidates. Slates that achieve the 5% threshold are coloured green.

In between these two safe regions is the “slate canyon”, the large yellow area. If this fictional party ran a slate containing anywhere from 38 to 291 of its best candidates, it could not possibly reach either threshold for expense reimbursement. For these slates, the average support for candidates would be below 5% and the total support across all ridings would be below 2%.

Of course, in real life it’s impossible to know the exact performance of candidates in advance. (Remember, this graph was constructed using past election results.) Because of this, the edges of the slate canyon are not clearly defined. If you like, picture a steep slope with plenty of loose rock: it’s best to stay a few extra steps back. A truly “safe” candidate slate for this party might be limited to less than 30 candidates or more than 310.

Different Parties, Different Pressures

A party polling just above 2%, as in the example above, experiences the worst possible “slate canyon” scenario. The two cliffs are extremely small and the canyon between them is very wide. This party must choose between two very divergent nomination and campaign strategies if it wants to be certain of receiving the expense reimbursement.

Meanwhile, a party polling just above 5% experiences almost no pressure. It should achieve both thresholds easily, regardless of how it constructs its candidate slate. And for a party polling in double-digits, the expense reimbursement is all but guaranteed.

Effects of a Two-Tier System

What effect does this system have on the decisions of small parties? First, let’s put the size of the expense reimbursement in context.

Looking at five parties of different size across four recent elections (twenty campaigns total), election expense budgets were always equal to a year’s worth of donations or more&mdash;typically two to three years’ worth. (This makes sense: to maximize your impact in a four-year election cycle, you should save for three years and spend that money in the fourth.)

In other words, a 50% reimbursement on campaign expenses is equivalent to an extra year of fundraising in a four-year cycle. This is an amount that no party can afford to ignore.

As we saw above, parties at the “slate canyon” level of support (2-5%) are forced into one of two nomination strategies. Those nomination strategies don’t exist in a vacuum&mdash;they have an effect on candidate quality and on campaign strategy. Specifically:

  • A party aiming for the 2% country-wide votes threshold needs to nominate as many candidates as possible. This pressures the party to accept lower-quality candidates or to resort to “paper” candidates (non-campaigning candidates). This can result in more candidate gaffes or scandals, creating an amateurish impression. However, this nomination strategy does allow the party to have a presence in all ridings, helping future growth. The campaign platform for this party would focus on country-wide appeal and avoid regionalism, in the hopes of maximizing total votes across all ridings.
  • A party aiming for the 5% average support threshold needs to nominate its strongest candidates and exclude the rest. This avoids any problem of lower-quality or paper candidates, but at the same time, slows the party’s growth in ridings without candidates. Furthermore, the party’s candidates are more likely to be concentrated in a single region, or in ridings that share some attribute (e.g. urban ridings), which creates pressure for the party’s campaign platform to cater to those ridings, sacrificing country-wide appeal.

These strategic distortions are bad for voters, who would generally prefer to have a wide selection of parties that offer high-quality candidates and campaigns that cater to all regions of the country. The fact that parties below 5% are financially penalized, potentially losing the equivalent of a year’s worth of fundraising every election cycle, further limits the quality and variety of voters’ options at the ballot box.

Parties in the “slate canyon” level of support can attempt to compensate for these distortions. If they plan to run a full slate, they can begin recruitment earlier, allowing them to maintain a higher standard for candidates and rely on fewer paper candidates. If they plan to run a limited slate, they can work to establish riding associations or other presence in ridings without candidates, and commit to a policy development process that ensures a platform with country-wide appeal.


Campaign expense reimbursement rules for parties are given in section 444 of the Canada Elections Act ( The text of section 444 is below; comments in [italics and square brackets] are mine.

  • 444 (1) On receipt from a registered party of the documents referred to in subsection 437(1) [the party’s election expenses report], the Chief Electoral Officer shall provide the Receiver General with a certificate that sets out the amount that is the sum of 50% of the registered party’s election expenses, as set out in the return for its general election expenses, that were paid by its registered agents and 90% — to a maximum of $250,000 — of the registered party’s accessibility expenses, as set out in that return, that were paid by its registered agents, if
    • (a) the Chief Electoral Officer is satisfied — even despite any statement that the registered party’s auditor has included under paragraph 438(2)(d) in a report under subsection 438(1) [a statement from the party’s auditor that the party may not have complied with Canada Elections Act financial administration rules] — that the registered party and its chief agent have complied with the requirements of sections 437 to 443 [various rule for submitting expense reports and audited statements];
    • (b) the auditor’s report does not include a statement referred to in any of paragraphs 438(2)(a) to (c) [a report of faulty or dishonest financial records, a different offence than failing to comply with financial reporting rules]; and
    • (c) candidates endorsed by the registered party received at least
      • (i) 2% of the number of valid votes cast at the election, or
      • (ii) 5% of the number of valid votes cast in the electoral districts in which the registered party endorsed a candidate.

To summarize the above in plain language: a party that has complied with expense reporting rules to the satisfaction of the Chief Electoral Officer (even if the party’s auditor had previously raised concerns), and which has not been reported for faulty or dishonest accounting by its auditor, and which has reached either of the two vote thresholds, is entitled to the reimbursement of 50% of its campaign expenses, as well as 90% of accessibility expenses up to $250,000.

Candidate Shortage a Serious Problem For Greens

With Prime Minister Trudeau triggering an election today (August 15th), the major parties are accelerating their efforts to nominate candidates by the deadline.

Green Party supporters should be alarmed by their party’s lack of progress in this area. Based on available evidence, the party is not on track to field a full slate of candidates this election, and could face serious financial and political consequences as a result.

Where Nominations Stand Today

The election date has been set for 20 September. Candidate nominations close 21 days before election day, so parties will have until 30 August to nominate their candidates. 338 candidates are required for a full slate. (Though the Bloc Québécois only targets Québec ridings, of which there are 78.)

With these facts in mind, the graph below shows the nomination progress of the major federal parties as of 14 August. The dashed lines are straight-line projections, based on each party’s pace of nominations over recent days. The two vertical lines indicate the election call and the close of nominations. (Click or tap the graph to view full size.)

Graph showing candidate nominations for major federal parties, from early July through to 14 August, followed by straight-line projections out to 30 August, the close of nominations.
Nominated candidates for major federal parties, as of 14 August.
Data as reported by the parties (sources at end of post).

The established parties are all on pace to nominate a full slate, with the exception of the Green Party, which is on pace to complete only 245 nominations. (Data for the People’s Party is sparse, but suggests a ballpark projection of 190 candidates for this semi-established party.)

It may be tempting, based on this graph alone, for Green supporters to assume that simply “picking up the pace” will be sufficient to correct the problem. However, a look at additional data provides reasons for concern.

The Application Process

Little is known about the internal decisions driving the Green Party’s candidate selection process. Their recruitment campaign, “Time to Run”, was launched February 5th and called on supporters to help recruit a diverse slate of community leaders.

Since then, member mailouts and internal governance meetings have provided a glimpse into the number of applications received and the processing of those applications.

The graph below shows the number of applications received, applications approved, ridings with at least one approved application, and nominated candidates. (Click or tap the graph to view full size.)

Graph showing the number of received applications, approved applications, ridings with approved applications, and nominated candidates for the Green Party of Canada. Data begins in mid-June and runs to 14 August, with straight-line projections shown out to 30 August, the close of nominations.
Number of candidate applications received, applications approved, ridings with at least one approved application, and nominated candidates for the Green Party of Canada. (Sources at the end of the post.)

Note: A February 25 announcement indicated that the party had received “hundreds of expressions of interest”. I was unable to determine whether an “expression of interest” is the same as a candidate application, so I have excluded this data point.

As with the previous graph, we see that current nominations are less than half of a full slate. Extending from the final nomination tally on 14 August, the darker dashed line shows the projected pace of nominations, while the lighter dashed line shows, for comparison, the pace needed to reach a full slate.

The triangular points indicate the number of applications received. Announcements on 7 May and 14 June both referred to “hundreds” of applications; these have been represented by a single data point showing 200 applications on 14 June.

The Leader’s Newsletter of 9 July made a more specific claim of 238 applications, and a report to the 27 July Federal Council meeting made reference to 304 applications having been received.

Assuming the party continued to receive applications at the same pace, it would have received about 350 applications today, and would receive about 380 by the nomination deadline. (Both numbers are represented by the darker dashed line extending from the most recent application tally.)

At first glance, this might appear to be enough to nominate 338 candidates with room to spare. However, there are some important factors to consider.

Overlapping Applications

Some ridings, naturally, produce multiple candidate applications. When this happens, a nomination contest is held, and only one of the applicants ultimately becomes the candidate. For this reason, more than 338 applications are needed to nominate a full slate. But can we estimate how many?

One source indicated to me that in the 2015 election, and again in the 2019 election, the party received approximately 1000 applications. The party was able to field a full slate in both of those elections, so we can assume that 1000 is a safe upper boundary. This would give an applications-to-nominations ratio of 3-to-1. (In the graph above, the pace to reach 1000 applications is indicated by the lighter dashed line that extends after the most recent tally of applications.)

Additional statistics from the 27 July Council meeting suggest a lower boundary. Of the 304 applications that had been received by that time, about 230 had been approved. Those 230 had applied in about 210 ridings, which suggest that at least 10% of applications overlapped, giving an applications-to-nominations ratio of 1.1-to-1.

This is a pretty wide range. Thankfully, there’s one other piece of data that can help us estimate an applications-to-nominations ratio: the concentration of the party’s support base in previous elections. It’s logical to assume that the strongest ridings would produce multiple candidate applications, while the weakest ridings would fail to produce even one, and would need to be contested by “parachute” candidates.

The math is too complex to include in this post, but based on the party’s highly-concentrated support, a reasonable benchmark is that the party would need 680 genuine candidate applications, plus 40 paper candidates, to field a full slate. (Giving an applications-to-nominations ratio of about 2-to-1.)

Using this ratio, the Green Party’s projected 380 applications would only be enough to field around 230 candidates. (190 approved applicants, plus 40 paper candidates.) This is a very similar number to the 245 nominations that we projected in the first graph, based on the current pace of confirmed nominations.

Rejections and Withdrawals

Beyond the normal math of nomination contests, candidate totals are also affected by rejected applications and withdrawn candidacies. Both of these are present in every party, but in a healthy campaign, they affect only a small number of applicants.

Anecdotal reports indicate that starting in late July (following the Green party’s very public leadership crisis), an elevated number of candidate applications are being withdrawn, with at least one confirmed candidate also withdrawing. The number of rejected applications is also reportedly higher than in past years, while some prominent or “star” candidates never applied to begin with.

Even if rejections and withdrawals are at elevated levels, that would be unlikely to make a big difference to the total number of applications the party would need to nominate a full slate. For example, the target number might be 720 instead of 680.

However, aside from the direct effect on numbers, these are negative events that also affect morale and interest levels. This slows down the pace of applications, which is a problem given that the party already needs to pick up that pace.


Why does it matter so much if a party doesn’t run a full slate? To begin with, it affects the perception of the party’s legitimacy: a party unable to field candidates in all 338 ridings either does not represent the interests of the whole country, or is not a well-developed and well-run national organization.

However, for the Greens, there are far more serious consequences looming.

First is the percentage of the national vote. In the 2019 election, the Greens received 6.6% support with a full slate of candidates. If we assume the party had only run 240 candidates in 2019, and those candidates had received the same 6.6% support in the ridings where they ran, then the party’s share of the nation-wide vote, all ridings being equal, would have been only 4.7%. In other words, the national vote share would under-represent the party’s true support, influencing public perception of the party’s strength.

However, this is not 2019. The Greens are currently polling at only 4.8%. Even if we assume support is the same on election day (which history tells us is unlikely), running only 240 candidates would result in a national vote share of 3.4%, a twenty-year low. This would be an embarrassing step backward for the party; however, the national vote percentage affects more than just perception.

One of the criteria for inclusion in the Leader’s Debates, in both 2019 and 2021, was that the party must have obtained at least 4% support in the most recent election. Depending on the criteria used for the next election, having received a national vote share below 4% in 2021 could result in the Greens being excluded from those debates.

A second concern, also related to the national vote tally, is that federal parties that receive more than 2% of the national vote are entitled to receive a reimbursement for 50% of their election campaign expenses.

If today’s polling holds until election day, the Greens would receive more than 2% national support, even with a reduced candidate slate.

However, as of July, the Green Party has only $300,000 cash on hand for both campaigning and normal operations, leaving it incapable of mounting a national election campaign of any meaningful impact. This makes it almost certain that support will decrease during the campaign, as the Greens seem to “disappear” from the perception of the average voter. That, combined with a reduced candidate slate, could realistically pull the party below 2% national support.

Should that happen, the total amount of public funding the party would lose would be in the neighborhood of half a million dollars.

Worse still, one of the factors considered by banks when determining whether to give a campaign loan is whether the party can expect to receive the expense reimbursement. At this point, short of a fundraising miracle, a campaign loan is the only pathway to a serious national campaign for the Greens, so inability to access a loan could effectively shut down the campaign.


By reasonable measures, the Green Party is not on pace to nominate a full slate of candidates. Party supporters should be deeply troubled by the possible consequences.

Under Canada’s first-past-the-post system, parties on the edge of “established” status are in a fragile position. There’s a very real possibility that sluggish nominations could be the trigger that pushes the Green Party into a self-reinforcing death spiral: inability to secure a campaign loan, an incomplete candidate slate, doubly-reduced election results, future debate exclusion, and ultimately, a deep blow to the party’s perceived legitimacy and future prospects.

Given that the number of applications received thus far does not appear to be sufficient to nominate a full slate, this can’t be cured by speeding up administration. The Greens will have to find a way to attract more candidate applications, and quickly.


Weekly tallies of nominated candidates come primarily from Politico’s Corridors newsletter (which in turn obtained those numbers by asking the parties directly):

Additional sources for numbers of candidates and, for the GPC, numbers of applications:

Other sources:

Green Party of Canada Leadership Results: Updated Projection

This is a non-partisan projection of the outcome of the Green Party of Canada’s 2020 leadership race. This post updates the projection I posted on September 26th to account for last-minute data.


Annamie Paul wins the leadership in the 7th round,
defeating Dimitri Lascaris 53.6% to 46.4%.

Expected Round-by-Round Results

The table below shows the expected standings in each round, along with eliminations and vote transfers. Values are rounded to the nearest 0.1%.

Continue reading “Green Party of Canada Leadership Results: Updated Projection”

Green Party of Canada Leadership Results: A Projection

This is a non-partisan projection of the outcome of the Green Party of Canada’s 2020 leadership race. The projection is based on a statistical model, developed over the past year, that was designed specifically to project ranked-ballot leadership races in Canada. Using this model, 100,000 simulations were run and the results analyzed.


Official portrait of Annamie Paul.

Annamie Paul wins the leadership in the 7th round.
She defeats Dimitri Lascaris 64.5% to 35.5%.

Expected Round-by-Round Results

The table below shows the expected standings in each round, along with eliminations and vote transfers. Values are rounded to the nearest 0.1%.

Continue reading “Green Party of Canada Leadership Results: A Projection”

Singh’s “Points of Difference”: National Unity

(This post is the second in a series. Previously, I examined Singh’s claim about Abortion Rights.)

At the first Leader’s Debate on Sept 12th, 2019, NDP leader Jagmeet Singh proposed four “points of difference” between the NDP and the Greens. One of his claims was that the NDP has “a solid position when it comes down to national unity”, and by implication, that the Greens do not. Is this claim legitimate?

Continue reading “Singh’s “Points of Difference”: National Unity”

Are Singh’s “Four Points of Difference” Between the Greens and NDP Legitimate?

During the Maclean’s–CityTV leader’s debate on September 12th 2019, NDP leader Jagmeet Singh contrasted his party with the Greens by naming four alleged points of difference between the parties. Are those differences legitimate?

Continue reading “Are Singh’s “Four Points of Difference” Between the Greens and NDP Legitimate?”

PPC vs. Greens: An Update

My previous post asked whether the People’s Party of Canada had overtaken the Greens. In that post, I compared polling, membership levels, donations, and number of EDAs. Several months have passed since then. This post updates my analysis with the most current information.

Continue reading “PPC vs. Greens: An Update”

Have the People’s Party Overtaken the Greens?

Recent announcements from the People’s Party of Canada indicate that, in the three-and-a-half months since it formed, it may have already surpassed the Green Party of Canada to become Canada’s fourth-largest federal party.

Continue reading “Have the People’s Party Overtaken the Greens?”