How candidates use Twitter says a lot about how parties communicate in Canada

Federal candidates in the recent election campaign mostly amplified their party leaders and mainstream news outlets on Twitter - which is exactly how the parties like it.

Candidates tended to echo the messages sent by leaders and official party accounts

CBC monitored candidates' retweets since May to find evidence of trolls spreading misinformation at the highest political level. (Matt Rourke/Associated Press)

Canadian parties are known to be highly disciplined in their communications. Twitter is no exception. 

From May until election day, federal candidates mostly retweeted messages from party leaders and official party accounts, a CBC analysis shows. Sometimes, they retweeted news accounts.

And that's exactly what the parties want them to do.

"The safest thing is to just do whatever leader is doing," said Alex Marland, a political science professor at Memorial University in Newfoundland. For the past year he's been writing a book about party messaging strategies. He said his conclusion is crystal clear: parties are very strict when it comes to what they allow candidates to say in public.

CBC collected retweets from 1,185 candidates and former MPs since May, a total of 300,000 retweets. (Regular tweets or tweets that mention another user were ignored.)

The goal was to see if candidates would get baited by online influence campaigns into retweeting and amplifying disinformation. But there was little evidence of that happening in the major established parties. The almost military discipline of Canadian politicians served as a bulwark against passive dissemination of misleading or false information.

"MPs are given material all the time to amplify. And if they stray away from leader's message, [it] could be problematic," Marland said.

"Even if they like a post on Instagram, it could get them into trouble, especially if it crosses party lines."

Toeing the party line

The diagram below is a representation of all the retweets collected. Each circle is a Twitter account. Those with yellow labels are candidates. White labels are everyone else.

The closer nodes are to each other, the more connections they share — the more they retweet each other.

Click on image to expand

Those coloured clusters exist because candidates retweet their fellow party members and other like-minded Twitter uses.

There's little overlap between party retweets. Some users, like Swedish climate activist Greta Thunberg, appear between the NDP and Green Party zones because she's retweeted often by candidates of both parties.

The People's Party of Canada (purple) and the Bloc Québécois (light blue) are more detached, which suggests silos of communication: there are few social media interactions between PPC and Bloc candidates and candidates for other parties.

But with the PPC, this lack of interaction is one-sided. PPC candidates frequently retweeted other parties to challenge them or to mock claims made by them. PPC Leader Maxime Bernier retweeted other leaders 45 times between May and October 9 — far more than other leaders did.

The other parties did not reciprocate. PPC candidates were retweeted the least by other party candidates.

If we look only at the number of times candidates retweeted other candidates, the PPC was the least retweeted of the federal parties apart from the Bloc Québécois.

This chart only includes retweets of other candidates. Retweets of other Twitter users were not counted.

In terms of retweets, the most insular party was the Liberal Party. The vast majority of retweets by Liberal candidates — 99.2 per cent — were of other Liberal candidates.

"Most candidates are aware that their success in the polls is related to the party leader's campaign," said Tamara Small, associate professor of political science at the University of Guelph.

Candidates retweet known news outlets

At the centre of the network diagram above is a cluster of news outlets' and journalists' Twitter accounts. This shows that candidates for the four main parties retweeted genuine news in high volumes.

The four most retweeted accounts were for the National Post, CBC's Power & Politics, CBC News and the Toronto Star. But they were not equally shared by parties. When broken down, we can see which news organizations were favoured by candidates.

That's to be expected, Small said.

"If a leader tweets a news story about that leader, candidates are making sure they amplify the national party message," she said.

Once again, the People's Party stands out: their candidates' favourite publication is The Post Millennial, which has a history of unreliable reporting mixed with advocacy.

To Small, this suggests that the newly-formed PPC doesn't have the same stringent policies on candidate behaviour as the older ones.

"I don't imagine Maxime Bernier puts candidates through the same kind of training as the Liberal Party," she said.

"This is in line with the party brand that claims to be real people speaking the real truth. It would be hypocritical of Bernier to have the same kind of message control."

Influence of bots on candidates was practically nil

Many feared that foreign trolls would try to interfere in the federal election by flooding social media with disinformation and divisive content to polarize Canadians.

Last year, CBC found evidence of Twitter accounts linked to the Russian and Iranian governments that appeared to want to provoke Canadians on divisive issues like immigration and pipelines. Throughout the campaign, researchers argued that trending hashtags like #TrudeauMustGo may have gone viral because of organized trolls.

But the influence of "bots" — automated accounts that could be run by troll farms — was almost non-existent, at least among the candidates themselves.

CBC used a service called Botometer, developed by Indiana University, to gauge how often candidates were retweeting potentially malicious accounts. Botometer analyzes an account's recent activity to estimate how much it behaves like a Twitter bot.

If we bring these suspected bot accounts into view, we see that they are mostly marginal — accounts that are retweeted only a few times by few candidates. (See the red dots in the network diagram below.)

Click on image to expand.

Not all bots are malicious. Some are useful, such as the bots that tweet alerts about earthquakes. And some people online can be falsely rated as bots because, by coincidence, they tweet in patterns similar to bots.

An account highly retweeted by Green Party candidates belongs to Paul Dawson, a self-described "climate change influencer" who mostly tweets news stories about the environment. He has a high bot score, despite apparently being run by a real person.

Percy Dastur, a candidate for the PPC, also was given a high bot score, despite having a fairly standard Twitter stream.

CBC identified just over 23 of these accounts that had troll-like qualities (five of those were later suspended by Twitter for violating the site's terms of use).

For Fenwick McKelvey, an assistant professor at Concordia University in Montreal who researches political communication online, this research confirms his view that the effects of trolls and bots on the campaign was overstated.

"Twelve per cent of Canadians use Twitter for news, so other than journalists amplifying the story by reporting on social media as public opinion, I don't see any evidence of this being effective," he told CBC.


Every night since mid-May, CBC scanned Twitter for retweets by former MPs and candidates using the Twitter API. Candidates' Twitter accounts were gathered from the party websites once a week. It's possible that there are candidates with Twitter accounts that were not listed on the party websites.

By October 20, 1,152 Twitter accounts were being monitored.

For every retweet found, CBC ran the retweeted account through the Botometer API and saved the bot scores for each one. Botometer scores were refreshed once a week.

CBC reporters manually looked through accounts with high bot scores for telltale signs of automation: user names with a long string of numbers or random words that could have been created by a computer, and Twitter accounts that mostly retweeted strongly partisan content. (For example: Carrie12138862, Richard74195773, WinstonCovfefeB.)

Add some “good” to your morning and evening.

A variety of newsletters you'll love, delivered straight to you.

Sign up now