Colors matter. As events unfold around the world, journalists focus our attention on the people, places, and things by creating scenes, much as film directors do. It isn’t just a man waving to a crowd from an airplane. It is a man in a “dark blue suit”, a “crisp white shirt” and a “bold red tie”. It isn’t just a crowd of women marching in the street. It is a sea of “pointy-eared pink hats”.

Looking back on 2017, what did our world look like in color? We analyzed over 30 million English-language news documents to find out.

RESULTS:

• 1 out of 5 stories in 2017 was colorful

• 7m out of 33m news articles used at least 1 of the 10 most common colors to describe something

• Grouping the 1000 color words of 2017 by category reveals our visual attention budget:

• 31% of the color words describe people’s clothing or body

• 15% are food

• 10% are species of plant or animal

• 10% are human-made objects

The rising colors of 2017

---

white supremacists

---

• 3-fold increase in mentions compared to last year

• Among the largest news narratives of 2017: Over 20k articles published by 2,629 news outlets grouped into 122 events. The stories mention over 1300 different people with 2,368 attributed quotes

• The first large pulse of mentions were driven by events related to the 11 August rally in Charlottesville, North Carolina, though earlier events were significant such as discussions of white supremacists recruiting on US college campuses

• Growth continued over the latter half of 2017 such as the rallies in Tennessee and policy shifts at Twitter

pink hats

---

• 17-fold increase in mentions compared to last year

• Largely location-driven stories in 2,491 unique news stories grouped into 69 events

• Dominated by the massive event of the women protest marches in Washington D.C. on 21 January

• Sustained coverage of follow-up marches in Boston and elsewhere, as well as continuing commentary

green cards

---

• 2.4-fold increase in mentions compared to last year

• Another very large news narrative of 2017: Over 20k articles published by 2,961 news outlets grouped into 84 events. The stories mention over 1000 different people with 1,571 attributed quotes

• Coverage of “green cards” was equally distributed across the year, from the Republican party dissent and Trump’s travel ban to scandals over green card-hawking and threats to terminate the diversity lottery

blue whales

---

• 3.1-fold increase in mentions compared to last year

• Surprisingly large news narrative for an animal (because it’s not just about the animal): 8k articles published by 2,173 news outlets grouped into 132 events. The stories mention 714 different people with 569 attributed quotes.

• News coverage of “blue whales” was divided roughly equally between 2 topics

• Coverage of the plural form (“blue whales”) was largely driven by events related to whales, including a new blue whale skeleton at London’s Natural History Museum and scientific progress in our understanding the evolution of massive body size and the discovery that blue whales have handedness. Average sentiment scores for these events were highly positive.

• In stark contrast, sentiment of coverage of the singular form (“blue whale”) was extremely negative. It was driven by coverage of an online game (called “blue whale”) involving teen suicides in the US and India.

red pandas

---

• 2.9-fold increase in mentions compared to last year

• Sneaking into the top-100 of this crowded color space (red carpets, red tape, red shirts) is the endangered red panda, an “adorable” and “extremely photogenic” cat-sized relative of the better-known giant panda. Nearly 2000 news outlets covered 48 events related to 617 locations around the world.

• Most of the coverage of red pandas was related to events at zoos, such as an injury to a red panda cub in Australia, a red panda on the loose in Virginia, and the birth of two red panda cubs in Denver.

• The exception is the launch of a Japanese cartoon character named Aggretsuko, a red panda with an office job.

The falling colors of 2017

---

yello fever

---

• 58% less coverage compared to last year

• Coverage of yellow fever in 2016 was dominated by a major outbreak in Africa and a successful large-scale vaccination campaign

• Coverage this year was dominated by a proposal to cut funding for the US Centers for Disease Control and a plan by one of Google’s sister companies to help fight the disease

black lives

---

• 49% less coverage compared to last year

• Coverage of “black lives” is almost entirely dominated by references to the grassroots movement that began as #BlackLivesMatter on Twitter

• Coverage crested last year, driven by the kneeling football protests and then spreading internationally

black holes

---

• In spite of this year’s Nobel Prize in physics going to the experimental confirmation of the existence of gravity waves from colliding black holes, journalists wrote articles mentioning black holes 13% less often this year compared to 2016.

• The dip in coverage is due in part to the comparatively massive coverage of the LIGO gravity wave experiment occurring last year

• This was the largest science story of the year. However…

red planet

---

• …continuing the trend of deprioritized science coverage in this unusually political (and polarized) year, journalists wrote 29% fewer stories about Mars (aka the “red planet”) compared to 2016.

• Elon Musk’s 2016 announcements of plans for the red planet feel like ancient history compared to this year’s news about “red states”

orange jumpsuits

---

• Although the number of incarcerated people in the US fell slightly between 2016 and 2017, it is an unlikely explanation for the 21% drop this year in news articles mentioning orange jumpsuits

• Intriguingly, 2017 saw a 229% increase in news about purple jumpsuits, though this was driven almost entirely by Harry Styles

Data | Tools:

• 33.5 million news articles published between 1 Jan and 1 Dec 2017 by 38,860 English-language media outlets

• Analysis done entirely in Python with Primer and spaCy

Method:

• Extract all sentences with one of these 10 common colors: white, black, red, green, blue, pink, brown, orange, purple, and yellow

• NLP (tokenization, lemmatization) to bin words independent of grammatical number (“hat” with “hats”, “child” with “children”, etc.) and surface the most common form

• Final count by document, i.e. 3 mentions of “red tie” within a news article equals just 1 mention for “red tie”

• Named entities excluded (only lowercase and contiguous color-word bigrams included for analysis)

• Words grouped into categories using word vectors, then manually tidied

• Change in color word frequencies based on comparison of 2017 (1 Jan - 1 Dec) vs. 2016 (1 Jan - 1 Dec)

• Deep-dive analysis based on Primer to reveal the events that drove the change

• Excellent collaboration with FoamTree to visualize the data – Dziękuję Ci!