All work

Exploratory Analysis · 2026

The Speed of News

How long the news of eleven events took to reach American newspapers, and later news websites, between 1799 and 2023.

Data & tools
Chronicling America, GDELT, Python, D3, Canvas
Format
Interactive piece

Loading visualization…

Interactive piece Open full screen (opens in a new tab)

About this project

The question

When George Washington died in 1799, half of the newspapers that reported it needed 18 days to do so. When an earthquake hit Turkey and Syria in 2023, half of the news websites that covered it had published within 7 hours. I followed eleven events across four eras to see how that gap changed over time.

Reading the plumes

Each event is drawn as a plume of lines rising from the moment it happened. Every line is one newspaper or one news website, and its height is the time the news took to reach it, on a scale from one minute to four months. Hovering over a line shows the paper or site, where it is, and the passage that carried the news.

The four eras are horse and sail, the telegraph, wireless, radio and television, and the web. I did not measure the years between the eras, so the chart shows empty gaps between them.

Where the data comes from

For the eight events from 1799 to 1963, I searched the Library of Congress's Chronicling America for the words a report would use, and then read each paper's pages in date order using the Library's OCR text. I placed each paper at the first page that reports the event. Because a paper carries a date but no time, each arrival is a window covering the day printed, and the summary figures use the end of that day.

From 1845 on, the half-way figures count only daily papers, because a weekly printed the news on its next issue day no matter how quickly the news had arrived. The weekly papers are still drawn, but fainter.

For the three web-era events, GDELT supplies the list of articles. GDELT records the time its crawler found each article, which can be hours after publication, so I used each site's own publication time wherever the page provides one. The half-way figures for the web count only those sites, and the remaining sites are drawn fainter.

Where the first pass was wrong

My first search for the Treaty of Ghent matched a false rumor and mentions of other treaties. The treaty did not reach America until February 1815, so the search now starts then. Rumors of Andrew Jackson's death appeared in Eastern papers days before he died and were denied, so I skip matches that appear near phrases such as "without foundation". For Pearl Harbor and Lincoln, several papers dated the day of the event used the name without reporting the event, either because they were printed before the attack or because they used the words in another sense, so I placed those papers at their first real report or left them out.

What to keep in mind

  • The eras are samples, and newspapers and websites are different kinds of source, so the comparison is between eras and does not form one continuous line.
  • Chronicling America contains only papers that libraries have digitized, which skew toward small-town weeklies and become sparse in the 1960s.
  • OCR misses words on faded pages, so some papers printed the news earlier than the chart shows, and the first report I found is never earlier than the true one.

Contact

What are you trying to
make sense of?

Get in touch