Skip to content
Join our Newsletter

The tweets of #elxn43: Tracking the negativity

An analysis of tweets from the party leader accounts provides a day-by-day look at the the ups and downs of campaign — and highlights which parties went negative and when.
Green Party leader Elizabeth May, left, responds to a question as Liberal leader Justin Trudeau, left to right, Conservative leader Andrew Scheer, People's Party of Canada leader Maxime Bernier, Bloc Quebecois leader Yves-Francois Blanchet and NDP leader Jagmeet Singh look on during the Federal leaders debate in Gatineau, Que. on Monday, October 7, 2019.

The Vancouver Sun analyzed more 5,500 tweets from the official Twitter accounts of party leaders since the start of the election campaign, scoring each tweet as positive or negative based on word choice.

Two periods stood out as the most negative for all candidates: the period following the discovery of Trudeau’s brown face photos, which saw increased negativity from all accounts, and in the days ahead of the two all-leaders debates, when accounts were used to attack other party leaders and their policies.

Overall, though, most tweets were classified as slightly positive or neutral — likely because the focus of the leader’s accounts was often on promoting campaign events or policy announcements rather than engaging with the public.

“It’s rare that we see party leaders authentically engage with voters,” said Anatoliy Gruzd, director of research at Ryerson University’s Social Media Lab.

“They’re busy on the campaign trail and they have a whole social media team managing the account so sometimes its hard to come off as authentic.”

At the same time Twitter is becoming a risky place for politicians, Gruzd said, because trolls and hate messages tend to surge following posts. But using Twitter to broadcast policy announcement and campaign updates in real time makes Twitter particularly appealing to politicians, he said.

“Now they don’t need to wait until the next morning to set up a press conference.”


Tweets from the official accounts of party leaders were analyzed starting on the day the election was called. Each tweet was scored using an Afinn model for sentiment analysis, which calculates a positive or negative value for each word in the tweet. Daily scores were calculated by taking the average from each day’s tweets, excluding retweets and French language tweets. Yves-Francois Blanchet published very few tweets in English and was dropped from the analysis.