Skip to content

Don't blame the election on fake news. Blame it on the media.

Authors: Duncan J Watts, David M Rothschild
Publication: Columbia Journalism Review, December 5, 2017

TL;DR

A critical analysis arguing that while fake news received disproportionate blame for Trump's 2016 victory, the mainstream media's coverage patterns—emphasizing scandals over policy—played a more significant role in shaping voter knowledge and behavior. Examining New York Times coverage, the authors demonstrate that scandal-focused reporting (particularly on Clinton's emails) vastly outnumbered policy coverage, potentially influencing voters more than the widely-cited fake news.

Key claims

  • Fake news impact is minimal: Fake news accounted for roughly 0.1% of Facebook's daily ad revenue in Q4 2016, and the top 20 fake news stories represented only 0.006% of user actions despite generating 8.7 million engagements.
  • Scale matters: While 126 million Facebook users saw Russian propaganda, Facebook users were exposed to 11 trillion posts (87,000 times more), and Russian tweets represented less than 0.75% of all election-related tweets.
  • Persuasion likely low: Research by Allcott and Gentzkow estimates the average American read only one or several fake news articles, with fake news potentially shifting vote share by hundredths of a percentage point—30 times less than Trump's actual margin of victory in key states.
  • Mainstream media dominated: The Washington Post published 50,000+ stories over 18 months; New York Times, CNN, and HuffPost each published 30,000+, vastly exceeding fake news in volume and engagement.
  • Scandal over policy: Analysis of New York Times front-page coverage (September 1 – November 8, 2016) found that scandal/personal stories outnumbered policy stories 3.4:1 on front pages and 4.2:1 overall, with 65,000 sentences about Clinton's email scandals compared to only 40,000 about all Trump scandals combined.
  • Email cascade: The Times ran 10 front-page stories about Clinton's emails in six days (October 29 – November 3), matching the number of all policy-focused articles across 69 days of campaign coverage.

Sources cited in the article

The article references studies and analyses including: - Hunt Allcott and Matthew Gentzkow's economics research on fake news exposure and persuasion - Harvard and MIT researchers' work on partisan media ecosystems (referenced in a CJR article about Breitbart) - Harvard/MIT MediaCloud report on media ecosystem dominance - BuzzFeed's viral fake news analysis - Pew Research Center polling on conspiracy belief - Shorensteiner Center research on news coverage of the 2016 election

Connections

Notes

This article is significant for several reasons. First, it provides quantitative grounding for the claim that fake news's direct impact on the 2016 election was likely negligible—not through dismissal, but through careful measurement and contextualization. Second, it makes a stronger and less-discussed argument: that mainstream journalism itself, through editorial choices about what to emphasize (scandals, horse-race coverage), shaped the electorate's understanding of the candidates' actual policy differences more profoundly than fake news did. The detailed analysis of New York Times coverage is particularly valuable for showing the disparity between scandal and policy coverage. The authors note their findings are typical of mainstream journalism broadly, not unique to the Times. A key weakness is that the article stops short of proposing solutions beyond acknowledging that news organizations should improve their own coverage. The conclusion—that fixing the information ecosystem requires improving "real news" as much as stopping fake news—remains provocative but underspecified about implementation.