How Negative is Political News Coverage?

Updated: 11-17-2017

This is a test-run of a new feature exploring how politics is covered by the mainstream U.S. news media.

Specifically, I’m undertaking an effort to estimate how negative – or positive – political news reporting is from major media outlets. At the moment, my analysis focuses solely on CNN and Fox News. CNN is included because they are a favorite target of the President, who frequently labels them “fake news,” and Fox News is included because they are generally regarded as having a conservative political slant, suggesting their coverage of the Trump administration might generally be more favorable than that of other news outlets.

The graphic below shows the average sentiment conveyed by CNN and Fox News between Saturday March 4, 2017 and Friday November 17, 2017.

Sentiment is defined here as the average number of positive words per article each day (words like “love,” “respect,” and “admire,” for example) minus the average number of negative words per article each day (words like “hate,” “terror,” and “disgrace,” for example).

Over the course of the past week, both CNN and Fox News used more negative words than positive words in their political news coverage. However, CNN was slightly less negative than Fox News. During the past seven days, the average political news story at CNN contained 5.39 more negative words than positive words. At Fox News, however, the average political news story contained 11.18 more negative words than positive words.

Over the last 259 days, the average political news article from CNN included 3.18 more negative words than positive words, whereas over at Fox News the average article included 4.29 more negative words than positive words.

The graphic below shows the relationship between the average sentiments conveyed by CNN and Fox News.

At the moment, there is a only a slight positive correlation (r = 0.27) between sentiments. This means, for example, that on days when political coverage at Fox News tended to be fairly negative (when the average news story included either many negative words or few positive words), political coverage over at CNN also tended to fairly negative.

However, this positive relationship appears to be driven, in large part, by just a few isolated and extreme events, such as the U.S. missile strike against Syria, which occurred on April 6, 2017. The following day on April 7, the average political news story at CNN contained 16.16 more negative words than positive words, and over at Fox News the average political news story contained 10.6 more negative words than positive words.


What were the Most Positive & Negative News Stories on Friday November 17, 2017?

Least Negative Stories:

CNN: DACA deal state of play (15 Negative Words vs. 30 Positive words)

Fox News: Jared Kushner’s ties to the White House, link to the Russia investigation (18 Negative words vs. 33 Positive words)

Most Negative Stories:

CNN: Trump blasts Franken, but stays silent on Moore (30 Negative words vs. 4 Positive words)

Fox News: Trump’s nicknames for rivals, from ‘Rocket Man’ to ‘Crooked Hillary’ (98 Negative words vs. 17 Positive words)


What Words Appeared Most Frequently in the News on Friday November 17, 2017?



What was Happening in the News this Week?

Below is a list of the top headlines from each of the past 7 days, starting with Saturday November 11, 2017 and ending with Friday November 17, 2017. The list comes from Memeorandum, a website that employs an algorithm to auto-generate a summary of popular news stories throughout the day. Each article was featured as the top story as of 1:00 PM EST on each respective day.

  1. Saturday, November 11, 2017: Trump says he believes Putin’s election meddling denials, CNN
  2. Sunday, November 12, 2017: Everyone Knew Roy Moore Dated High School Girls, Says Former Colleague, HuffPost
  3. Monday, November 13, 2017: Ignore the spin. Trump and the GOP have made a devil’s bargain with Roy Moore., Washington Post
  4. Tuesday, November 14, 2017: Locals Were Troubled by Roy Moore’s Interactions with Teen Girls at the Gadsden Mall, New Yorker
  5. Wednesday, November 15, 2017: LISTEN: Curious Robocall Seeks ‘Damaging’ Information on Moore, WKRG-TV
  6. Thursday, November 16, 2017: Senator Al Franken Kissed and Groped Me Without My Consent, And There’s Nothing Funny About It, KABC-AM
  7. Friday, November 17, 2017: I’m a feminist. I study rape culture. And I don’t want Al Franken to resign., Washington Post


A Little About the Methods

The findings presented here are based on a sentiment analysis of 7,074 articles from CNN (scraped from CNN’s politics RSS feed) and 2,538 articles from Fox News (scraped from Fox News’ politics RSS feed). The difference in the number of articles between the two news organizations is due to the fact that CNN offers 30 articles in their news feed at any given time, whereas Fox News offers only 10 articles in their news feed at any given time. Because the analysis of Fox News’ political coverage is based on a smaller sample size of articles, we should expect to see a greater number of extreme values – conveying either positive sentiment or negative sentiment – for Fox News relative to CNN.

Articles were scraped from each news feed at approximately the same time each day, roughly between 12:00 pm EST and 4:00 pm EST.

Text mining and sentiment analysis were performed using the tidytext package in R in conjunction with Bing Liu’s Opinion Lexicon.

A few words that were not intended to express either positive or negative sentiment, but which were flagged by the opinion lexicon as expressing sentiment, were removed from the corpus prior to analysis. Some examples of words that were removed for this reason include: “trump” and “right” – marked in the opinion lexicon as expressing positive sentiment – and “conservative” and “vice” – marked in the opinion lexicon as expressing negative sentiment.

Of course, these results should not be viewed as anything more than mere estimates of the average sentiments conveyed by CNN and Fox News. Moreover, these results most likely DO NOT reflect partisan bias in the media. For a listing of some important caveats and limitations to keep in mind regarding this analysis, and for a discussion of what these findings might – and probably do not – tell us about political news coverage, check out this post.



Brian Kurilla is a psychological scientist with a Ph.D. in cognitive psychology. You can follow Brian on Twitter @briankurilla