Vol. IFrozen 2026-09-11Git 2dfc1b5*220,221 articles · 1900–1963

GenderNews Atlas

Sections

74% of the people these newspapers name carry no textual sign of their gender.

Measured on 1,062,339 people in 220,221 articles sampled from 1900 to 1963. Before anyone asks how the share of women changed, that number decides what the share can mean. This atlas tracks both, and tests whether the answer survives a change of method.

See what survives the robustness grid

The question

How did the representation of women and men in American newspapers change between 1900 and 1963, and how much of any answer is an artefact of the NLP method used to measure it?

The second half of the question is the point. A trend line from one pipeline says little on its own. Here every headline estimate is re-measured with different rules for who counts as a person, what counts as evidence of gender, and what counts as a role. Estimates are shown with the corpus's changing composition held fixed and with it left free.

What holds, and what does not

Each hypothesis was written down before any measurement was run. A status of supported means the pre-registered criterion holds in the primary specification and the sign survives every cell of the robustness grid; tentative means only the first; unsupported means neither. Statuses are computed by scripts/claims.py.

  • The share of gender-signalled people who are signalled as women increases, 1900-1963.

    tentative · H1 · evidence

  • Women's share of authority roles is below their share of all people in every period, and the gap narrows.

    unsupported · H2 · evidence

  • Adjusting for topic and newspaper changes the 1900-1963 trend by at least 25% of the raw slope.

    supported · H3 · evidence

  • Role-measurement methods produce materially different trend effect sizes (ratio > 2 or sign disagreement for at least one headline role).

    supported · H4 · evidence

  • The lexical method (A) yields a larger absolute trend than the context-aware methods (B, D).

    tentative · H5 · evidence

  • The UNKNOWN share changes over time, and the women's-share trend is sensitive to how UNKNOWN is handled.

    tentative · H6 · evidence

  • Among attributed quotations, women's share of speakers is lower than their share of gender-signalled people.

    supported · H7 · evidence

Women's share of gender-signalled people, by gender-evidence ruleShaded band: 95% article-cluster bootstrap CI for the primary rule.
Table view
Yearprimary: honorific + pronounhonorific only+ gendered nouns
190043.9% (41.6%–46.2%) · n=10,17956.6% · n=7,19043.0% · n=10,680
190347.2% (44.4%–50.2%) · n=9,71061.8% · n=6,69546.2% · n=10,245
190651.2% (49.2%–53.1%) · n=10,07667.2% · n=7,06750.0% · n=10,628
190951.9% (49.7%–53.9%) · n=10,40768.3% · n=7,41050.5% · n=10,995
191257.3% (55.4%–59.2%) · n=10,93273.3% · n=8,02355.9% · n=11,525
191561.7% (59.7%–63.7%) · n=10,75876.9% · n=8,06460.2% · n=11,341
191862.0% (60.0%–63.9%) · n=9,87777.0% · n=7,50960.0% · n=10,490
192164.4% (62.9%–65.9%) · n=12,64677.5% · n=9,94462.4% · n=13,382
192463.4% (61.6%–65.3%) · n=11,23880.0% · n=8,36661.7% · n=11,853
192767.8% (66.0%–69.5%) · n=12,59385.1% · n=9,43166.0% · n=13,240
193069.6% (68.1%–71.1%) · n=12,73885.2% · n=9,79167.2% · n=13,551
193367.5% (65.6%–69.2%) · n=13,60583.3% · n=10,37965.5% · n=14,473
193668.1% (66.4%–69.8%) · n=13,90883.7% · n=10,67565.9% · n=14,901
193970.4% (68.9%–72.0%) · n=15,00283.4% · n=12,18768.5% · n=15,879
194267.3% (65.6%–68.9%) · n=11,78583.5% · n=9,03564.7% · n=12,679
194565.2% (63.6%–66.8%) · n=11,20681.8% · n=8,47962.2% · n=12,430
194866.4% (64.9%–68.0%) · n=14,11777.0% · n=11,64164.3% · n=15,179
195167.3% (65.5%–68.8%) · n=14,25979.5% · n=11,63265.4% · n=15,303
195468.5% (67.1%–70.0%) · n=13,71480.6% · n=11,23166.5% · n=14,656
195771.5% (70.2%–72.7%) · n=15,21281.7% · n=12,84669.3% · n=16,264
196070.2% (68.5%–71.4%) · n=15,44580.4% · n=13,03367.8% · n=16,785
196365.0% (63.3%–66.5%) · n=12,30978.0% · n=9,76762.4% · n=13,636

AmericanStories (Dell et al. 2023, CC-BY-4.0), built on Chronicling America (Library of Congress). 10,000 sampled articles per year, 22 years.