Read That Social Media Swings Elections? That’s Probably Just Fake News

Facebook, Twitter and the internet more broadly have limited explanatory power for the state of politics today.
November 7, 2017 • Commentary
This article appeared in City A.M. on November 7, 2017.

It is often said that “a lie gets halfway around the world before the truth has a chance to get its boots on”.

The great and good all seem to agree that there’s a new boogeyman for fake news: the internet – or, more specifically, social media.

In the aftermath of the Brexit vote and the election of Donald Trump, the charge sheet for sites such as Facebook and Twitter is long. They are said to give a platform for wild conspiracy theories, facilitate the spreading of misinformation, and help create “echo chambers” or “filter bubbles”, whereby people only see news and views which back up their own opinions.

The Economist ran a front page story last week implying that social media was the nemesis of democracy. The BBC’s media editor wrote a post days beforehand asking “can democracy survive Facebook?”. In the US, Facebook has been in the firing line of Democratic politicians since it came to light that as many as 126m Americans may have seen Russian‐​sponsored ads during the Presidential election campaign.

Conventional wisdom increasingly suggests that social media is polarising our politics and even – in the US case – helping to swing elections.

But is there any evidence to justify these conclusions?

In a recent paper, the economists Levi Boxell, Matthew Gentkow and Jesse Shapiro collected data to test whether the internet was to blame for polarisation and could explain the shock Trump victory in 2016.

Their results might surprise you.

Constructing a measure of polarisation of American voters, they indeed find a significant increase over the past two decades, a time during which the internet and social media use has taken off.

If the internet itself was driving this trend, rather than something else, we would expect polarisation to be most intense among demographic groups with the most exposure to online information.

In fact, they find the complete opposite.

The increase in polarisation in the past 20 years is twice as large for over 65s compared to 18–39 year olds, despite the former obtaining information online and using social media at significantly lower rates.

Similar results are found when controlling for other potential demographic determinants of internet usage, such as education and race: the groups least likely to have internet access displayed greater increases in polarisation.

They found no support either for the idea that the internet was critical to Trump’s success. Using three measures of internet exposure, Trump did as well or better than his Republican predecessors among groups with low internet exposure, and as well or worse among groups with high internet exposure.

In two of the three measures, the 2016 election in fact saw the Republicans’ vote share among the low internet group exceed the high internet group for the first time.

Now, it could be that Trump would have done much worse among regular internet users absent developments in social media, or that news or information spread by younger cohorts gets transferred to older groups through other means where its polarising effects are amplified.

But on the face of it, there is no supporting evidence from this work that social media can be blamed for polarisation, or indeed credited with shaping the election.

This is important. Some commentators regularly condemn social media firms’ effect on politics, and even suggest that websites such as Facebook should be regulated as if they are public utilities.

While it might seem to make sense to blame new forms of media for political disruption, this evidence implies that online polarisation is reflective of societal trends, rather than social media itself driving the polarisation.

It is tempting for people upset with recent electoral outcomes to reach for boogeymen to explain the results. Social media and the internet are increasingly easy targets, especially given Trump’s regular Twitter use. But this recent paper suggests strongly that Facebook, Twitter and the internet more broadly have limited explanatory power for the state of politics today.

Perhaps, just as with other method of communication such as telephones, we should think of social media instead merely as a medium to express free speech.

As the economist Tyler Cowen has written, free speech itself often leads to the articulation of bad ideas that we may disagree with. It can reveal ugly undersides to society, and can encourage groupthink.

But attacking the mediums used for free speech makes no sense.

And we should not be surprised that the biggest attacks on Facebook come from the mainstream media, who stand to lose most from social media dominating the spread of information.

About the Author
Ryan Bourne

R. Evan Scharf Chair for the Public Understanding of Economics