The survival analysis dataset where the subject is the industry that reported its own decline: 24,242 newspapers across 313 country labels and 224 languages, founded and dissolved years making it survival-ready with Still_Published as the censoring indicator. The completed lifespans are brutal and real: the median paper that died lasted 6 years, while the longest, Sweden's Post- och Inrikes Tidningar, ran 364 years from 1645 to 2009, the world's oldest newspaper ending as a website. The recorded death wave peaks in the 1930s (1,090 dissolutions), not the internet era, and only 489 deaths are recorded since 2000, which is the dataset's honest caveat stated plainly: the graph records funerals slowly, so absence of a dissolution date means censored, not alive, and a 1617 Frischmann-Berichte whose own title ends in 1626 still carries Y. That is informative censoring, the exact complication survival courses warn about, shipped as a live example instead of a footnote. Dates are best-rank and year precision or better (founding on 20,571 rows, and the fix that made it so is in the scraper: The Guardian carries multiple founding statements and an arbitrary one said 1870), a paper cannot die before it is born by test, and anchors are checked by QID because names are not keys: the dataset holds eight Guardians, Charlottetown's among them. The country column preserves dead states as first-class: 2,504 papers filed under Austria-Hungary and 567 under Czechoslovakia, publishers outlived by their own bylines. The New York Times runs since 1851, News of the World closed at 168. Built with a 13-test acceptance suite.
24,242 rows in one CSV file, with 12 columns in total: Newspaper, Newspaper_ID, Country, Language, City, Founded_Year, Founded_Decade, Dissolved_Year, Still_Published, Lifespan_Years, Format, Newspaper_URL.
It is built by programmatically scraping wikidata.org, last pulled on 2026-10-08. Datasets are versioned; older versions stay downloadable.
Yes. Every CodeSights dataset is completely free as a CSV download; a free account is all it takes. Anyone can preview the data without signing in.
Yes. The exact Python scraper that built it is viewable on the dataset page by any signed-in member, so every number is reproducible.
Automated scraping leaves room for error and the underlying sources change over time, so no version is guaranteed accurate or complete. If a number matters, verify it against the original source.