Twitter study shows pro-Trump tweets swamped Clinton’s in Michigan

In this Election Day photo, Republican presidential candidate Donald Trump speaks at a campaign rally in Grand Rapids, Mich.
In this Election Day photo, Republican presidential candidate Donald Trump speaks at a campaign rally in Grand Rapids, Mich. AP

Voters in Michigan received nearly three times as many Twitter messages favoring Donald Trump in early November compared with tweets supporting Hillary Clinton, a British research team has found.

Among the politically related tweets that carried links to content during the period from Nov. 1-11, nearly as many shared “junk news” such as propaganda and information from ideologically extreme political sources as did those carrying links to professionally gathered news stories, the study from Oxford University reported.

“The problem is that voters in a state like Michigan were learning about conspiracies more than fact,” said Prof. Philip Howard of Oxford’s Internet Institute, who led the research.

Trump’s narrow victories in Michigan, Pennsylvania and Wisconsin, by a combined margin of fewer than 78,000 votes, enabled him to capture the White House.

The researchers from Oxford University, who have done multiple analyses of social media in last fall’s election, sought to take a snapshot of the potential impact of messages to voters in an era of social media expansion.

They released two analyses: One about the types of politically related social media in Michigan, which Trump won by about 10,700 votes, and a second examining the dimensions of fake news and other social media in Germany during its election.

In Michigan last fall, residents sent 138,686 politically related tweets during the seven days leading to the Nov. 8 election and for three ensuing days, according to Twitter data collected by the researchers. Of those 78,662, or 56.7 percent, were pro-Trump and 28,074 pro-Clinton, they said.

Of 24,291 Michigan tweets that shared links to political information, 5,615 directed recipients to “junk news,” such as propaganda and information from ideologically extreme or conspiratorial political sources, while 5,668 linked to professionally gathered news stories. A March 2016 survey by the nonpartisan Pew Research Center found that about 21 percent of Americans were using Twitter, a figure that surges to 36 percent for people ages 19 to 29.

“Electing a politician is one of the most important democratic exercises we’ve had,” Howard said in a phone interview. “We’ve fought several world wars to protect that right. Voters these days are making decisions with very poor quality information about the problems that we all have to deal with.”

Howard said the study tracked a couple of days after the election in an effort to gauge what normal Twitter activity would look like.

The analysis was released amid a broad federal investigation into a multi-pronged Russian cyber attack that U.S. intelligence agencies say was aimed at undermining public confidence in democracy, helping Trump and defeating Clinton. Russian President Vladimir Putin had blamed the former secretary of state for inciting riots against his regime.

A Senate subcommittee also is investigating alleged Russian interference in European elections, including through social media.

McClatchy reported last week that federal investigators are examining whether far-right news sites had any contact with operatives for Russia who designed robot-like automated programs, or “bots,” that sent Twitter and Facebook messages to millions of Americans to heighten readership of pro-Trump news on those sites.

Howard said that a third of the so-called junk news shared via Twitter in Michigan during the 10-day period came from five far-right news sites — Breitbart News, InfoWars, the Daily Caller, Liberty Alliance and Truthfeed.

Oxford’s Internet Institute, however, could not identify whether any content was shared via Russia-generated bots, he said.

The new study did find that in early November, bot traffic in Michigan accounted for only about 2 percent of the politically related Twitter messages, though the researchers said, “It is possible that some of these platforms are obscuring accounts that really are highly automated.”

The Oxford researchers cited “numerous examples of misinformation distributed online with the intention of misleading voters or simply earning a profit through scandalous headlines and evidence of manipulation of online public opinion.”

The World Economic Forum, a not-for-profit foundation that focuses on global issues, recently said that the growth of misinformation online was one of society’s top 10 perils, they said.

In Germany’s federal election in February, the researchers found that tweets related to the far-right presidential candidate, Albrecht Glaser, were large in comparison to his share of the vote. But bots were not a big factor, and the ratio of “professional news to junk news” shared on Twitter by Germans was 4-1, they said. However, they said bots had been used in Germany to spread “fake news.”

“Political actors and governments worldwide have begun using bots to manipulate public opinion, choke off debate and muddy political issues,” the report said.

Howard said he believes the analysis marks “the first time we have some numbers about how well the far right uses Twitter effectively.”

Greg Gordon: 202-383-6152, @greggordon2

David Goldstein: 202-383-6105, @GoldsteinDavidJ

Related stories from McClatchy DC