Can social media read a by-election before the vote?

Since 19 September Qalu has followed the Brantford–Brant South federal by-election in the open, without taking a side. This is what the conversation looked like a month before polling day.

posts, comments and articles
2,478
of it on Facebook
71%
records on the busiest day, 8 November
574
news articles, against 2,174 comments
54

The report

Thirty-eight slides, published on 9 November: the race as a whole, then each candidate in turn. The boards below are where the numbers come from.

  1. Cover: Brantford–Brant South by-election
  2. More negative than positive
  3. 84% of the talk came after 25 October
  4. Facebook leads, TikTok is most negative
  5. The appointment debate is the most negative topic
  6. Davis and Carrol take 37% of all records
  7. A Conservative seat, now contested
  8. Instagram and YouTube punch above their weight
  9. TikTok is negative for every candidate
  10. National politics outdrew the candidates
  11. News surged in the final two weeks
  12. News and public talk diverge
  13. Both big parties draw equally negative talk
  14. The appointment and living costs drive the negativity
  15. Only Carrol’s comments lean positive
  16. One name, then two, then four
  17. Attention is not approval
  18. Each candidate owned a different week
  19. Poilievre is named more than any local issue
  20. What the data supports, and what it cannot
  21. Party and leader names fill the vocabulary
  22. Six comments under one video hold 17% of all likes
  23. Kevin Davis, Conservative
  24. Kevin Davis: two waves of attention, both negative
  25. Kevin Davis: the appointment draws his sharpest criticism
  26. Kevin Davis: party words dominate
  27. Dave Carrol, Independent
  28. Dave Carrol: quiet until his independent run
  29. Dave Carrol: praise for the campaign, doubt about vote splitting
  30. Dave Carrol: good, right, and lawn signs
  31. Joy O’Donnell, Liberal
  32. Joy O’Donnell: visible only after 1 November
  33. Joy O’Donnell: the debate is about the party, not her
  34. Joy O’Donnell: “Vote Liberal” and “vote blue”
  35. Brad Ward, NDP
  36. Brad Ward: housing and seniors set the agenda
  37. Brad Ward: housing and homelessness are the words
  38. Four takeaways from the data
1 / 38 · Cover: Brantford–Brant South by-electionSwipe or use the arrows
The question

What is this race about, and where is it being argued?

A federal by-election is a small, local contest. National pollsters rarely survey a single riding, and local papers have thin newsrooms. Yet the people who live there talk about it every day, mostly online.

We wanted to know whether a monitored sample of that talk could show what the race is about, where it happens and when it heats up. We said so publicly at the start, and we will set what we saw against the result after the vote.

What we monitored

One riding, eleven sources, and a rule about place

The project searches news sites and ten social platforms for the riding’s names, the by-election and the candidates, from 18 September to polling day.

Candidates’ names are common names. On their own they would pull in strangers from across the country, so every record must also mention Brantford or Brant. Only seven records were thrown out as unrelated.

The search set-up of the Brantford–Brant South project in Qalu: the terms it looks for, the place each record must mention, and the sources searched
  1. A The place rule: a record counts only if it also says Brantford or Brant.
  2. B News sites plus ten platforms, from X and TikTok to Reddit and Bluesky.
The search behind the project, as it appears in Qalu.
What we found

A conversation held in the comments, on Facebook

Nine records in ten are comments: 2,174 of 2,478. The race is argued under posts, not in them. News sites supplied only 54 records.

The tone leans negative, as campaign talk usually does: 37% of classified records against 28% positive. Tone measures wording, not voting intention.

The overview in Qalu: 2,478 records tracked, 414K views, 52.2K engagements, 2.20M estimated reach, 28% positive and 37% negative
  1. A 304 posts and articles, 2,174 comments.
  2. B Negative outweighs positive. That is wording, not votes.
The project’s overview in Qalu, 19 September to 17 November. 9 in 10: records are comments, not posts.
Daily coverage in Qalu: almost nothing until late October, then a steady rise to a peak of 574 records on 8 November
  1. A 28 October: the race wakes up, 117 records in a day.
  2. B 8 November: 574 records, the busiest day so far.
Daily coverage, coloured by tone.
Platforms in Qalu: Facebook 1,762 records (71%), TikTok 256, Instagram 224, YouTube 121, LinkedIn 61, news sites 54
  1. A Facebook carries 71% of the records and 46% of the engagement.
  2. B Instagram and YouTube: fewer records, more views than Facebook.
Where the records came from. 71%: of the race is argued on Facebook.

What it adds up to

  1. The race lives on Facebook

    1,762 of the 2,478 records came from Facebook. A campaign that ignores it is absent from most of the local conversation.

  2. The talk began late

    For a month the riding was almost silent. It woke on 28 October and peaked on 8 November with 574 records.

  3. Candidates first, issues second

    Candidates and campaigns account for 1,102 records; local issues for 195; leaders’ visits for 174; voting procedures for 79; polls for 36.

Have a question like these?

Tell us what you need to follow. We will show you the search, the board and the report on your own subject.