Death gave a 2.4-fold year-1 rise in English Wikipedia views but no clear median change in years 2–4 against matched living people (1,517 deaths, 2016–2021); effects fell with age

Question: does the death of a public figure durably change attention to their English Wikipedia article? Earlier work compares views after death with the same person's views before death. Method: Wikidata humans who died 2016-07 to 2021-12 with >=12 sitelinks; daily user pageviews 2015-07 to 2026-09 from the Wikimedia API (3,350 deceased and 2,750 living series, random interleaved download with a fixed stopping rule). Each deceased person with >=20 views/day in days -365 to -31 was matched to 5 living people with the same occupation group and gender, a near birth year and the nearest pre-death views on the same calendar days. Effect = change in log mean views of the deceased minus that of the controls. Placebo: one control against 5 other living people. Results (n = 1,517): the year-1 median ratio was 2.36 (95% CI 2.26-2.46). The decay follows c + a t^-b with b = 1.51 (better than exponential models by AIC). For years 2-4 (pre-specified primary outcome) the median effect was +0.009 log units (-0.020 to +0.035); the placebo gave -0.004. A self-baseline without controls gave +3% to +4% per year, because the matched controls also increased. 9.0% of people were above the placebo 95th percentile and 8.6% below its 5th percentile, so death made long-run attention more dispersed. The effect decreased by 0.113 per 10 years of age at death and was higher for anglophone citizens (+0.114) and for arts than for leaders or sports. Mean year-2 effects by age band (<60: +0.35; 60-74: +0.11; 75-84: +0.06; >=85: -0.11) replicate SCI 2c5f9052, an independent study with a different sample. Effects in years 5-9 were negative (-0.04 to -0.13), but this depends on age matching: for the oldest people the controls were younger, and controls were selected on present-day Wikidata status. Limitations: selection on 2026 sitelinks in both arms, few well-matched controls above age 85, title-based counts, and bootstrap intervals that ignore control reuse.

collective attention · collective memory · Wikipedia pageviews · death · matched controls · difference-in-differences · power-law decay · placebo test · Wikidata · replication