roaddanger.org

Scraped · 2024 Netherlands Report

Only 2.4% of Dutch crash headlines tell the human story.

We collected every Dutch news article about traffic crashes published in 2024, kept the ones about a real crash with more than one party, and tested each headline.

Articles analysed
4,979
Crashes
3,387
People killed
340
People injured
2,927

The humanization test

Most headlines fail at the first question

The questions build on each other, like the Bechdel test: a headline only gets to the next question if it passed the one before.

  1. 1Mentions all involved crash partners
    40.8%
  2. 2As humans, not as transportation modes
    8%
  3. 3Written in active grammar
    2.4%

Beyond the headline: 25.7% of articles describe what happened to everyone involved, and just 5.2% place the crash in a larger pattern.

Over the year

Month after month, the same pattern

Where each month's headlines stopped passing the test.

  • Doesn't mention all parties
  • Mentions all parties…
  • …as humans…
  • …in active voice
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec

Rankings

How does your outlet score?

Outlets with at least 20 articles, ranked by how far their headlines get through the test. Online and print editions are listed separately.

  • Doesn't mention all parties
  • Mentions all parties…
  • …as humans…
  • …in active voice
  1. 1de limburger.nlonline
    223 art.
  2. 2BN De Stem.nlonline
    574 art.
  3. 3AD/Algemeen Dagblad.nlonline
    2,007 art.
  4. 4De Gelderlander.nlonline
    966 art.
  5. 5Eindhovens Dagblad.nlonline
    392 art.
  6. 6Brabants Dagblad.nlonline
    431 art.
  7. 7PZC.nlonline
    317 art.
  8. 8De Stentor.nlonline
    756 art.
  9. 9Tubantia.nlonline
    398 art.
  10. 10Noordhollands Dagblad.nlonline
    318 art.
  11. 11Dagblad De Limburger.nlonline
    28 art.
  12. 12Het Parool.nlonline
    25 art.
  13. 13HLN.beonline
    69 art.
  14. 14Leidsch Dagblad.nlonline
    103 art.
  15. 15De Gooi- en Eemlander.nlonline
    82 art.
  16. 16IJmuider Courant.nlonline
    115 art.
  17. 17Haarlems Dagblad.nlonline
    116 art.
  18. 18Brabants Dagbladprint
    57 art.
  19. 19De Telegraaf.nlonline
    165 art.
  20. 20Haarlems Dagbladprint
    35 art.
  21. 21IJmuider Courantprint
    21 art.
  22. 22nu.nlonline
    74 art.
  23. 23Leeuwarder Courantprint
    37 art.
  24. 24De Limburgerprint
    61 art.
  25. 25Noordhollands Dagbladprint
    214 art.
  26. 26AD/De Dordtenaarprint
    25 art.
  27. 27AD/Rivierenlandprint
    30 art.
  28. 28BN/DeStemprint
    129 art.
  29. 29Dagblad van het Noordenprint
    29 art.
  30. 30AD/Haagsche Courantprint
    124 art.
  31. 31AD/Rotterdams Dagbladprint
    102 art.
  32. 32AD/Groene Hartprint
    33 art.
  33. 33De Gelderlanderprint
    106 art.
  34. 34Provinciale Zeeuwse Courantprint
    39 art.
  35. 35De Stentorprint
    164 art.
  36. 36De Telegraafprint
    37 art.
  37. 37De Twentsche Courant Tubantiaprint
    65 art.
  38. 38Eindhovens Dagbladprint
    87 art.
  39. 39AD/Amersfoortse Courantprint
    55 art.
  40. 40AD/Utrechts Nieuwsbladprint
    48 art.

Matrix

Who kills who?

340 people killed. Rows: who was hurt. Columns: who they collided with.

UnknownPedestrian(Electric) kick scooterBicycleMotorcycleCarTaxiDelivery vanEmergency vehicleAgricultural vehicleBusTramTruckTrain
Unknown·············2
Pedestrian2··1222·3·113716
(Electric) kick scooter···111322··1·21
Bicycle1·1335239156·182
Motor scooter·····1········
Motorcycle2···37·3223·2·
Moped car·····1·1······
Car1····58·6·23·424
Delivery van·····1········
Bus·1············
Tram·1············
Truck·········1··1·
Train·1··········31

Map

The Crash Map

Every crash we could place. Dark rings are fatal crashes. Click a dot for the headline.

Method

From 56,785 articles to one corpus

Articles come from LexisNexis. Filtering, crash details and the headline test were done with an LLM and checked against volunteer labels from roaddanger.org.

  1. Raw LexisNexis export (2024 publication dates)
    56,785
  2. Articles with a headline
    56,783
  3. After removing automated police-feed notices
    22,622
  4. After removing exact title/date/outlet repeats
    22,163
  5. After scope filter (off-topic / single-sided / roundup)
    15,720
  6. After removing remaining automated notices
    14,886
  7. After merging identical syndicated copies
    9,855
  8. After merging near-identical copies
    9,266
  9. After crash located in the Netherlands
    6,721
  10. After passes secondary relevance check
    6,442
  11. After more than one party involved
    5,630
  12. Analysis corpus
    4,979

Help change how we talk about crashes.

For journalists and editors

Sign the pledge

Join the outlets committing to human-centered reporting

For everyone

Try the ReFrame tool

Paste any headline and rewrite it to tell the human story

For everyone who wants evidence

Explore the data

Dive into the data on how the media reports crashes