Births conceived during Ramadan in national monthly statistics: a test using the lunar drift

Ramadan moves through the Gregorian calendar by about 11 days a year, so a change in conceptions during the fast would show in births about nine months later, moving with the lunar date. This paper relates Ramadan-timed conception windows to national monthly births in the UN Demographic Statistics Database (table 55, by year of occurrence), with country-by-calendar-month and country-by-year fixed effects, so the exposure varies only through the lunar drift; a Ramadan reading needs nothing else affecting births to share that drift, which the planned placebo test questions. The analysis plan, published with this paper, cannot be shown to precede any inspection of the data. In the planned primary analysis of 17 Muslim-majority countries (471 complete country-years, 1969-2025; Iran excluded), a birth month whose entire conception window falls in Ramadan has 1.6% fewer daily births (95% wild cluster bootstrap interval -4.1% to +0.8%; p = 0.17), null under the planned rule. The planned placebo set, 54 countries with a Muslim share under 5%, gives -0.7% (-1.25% to -0.15%; p = 0.01), so the placebo test failed. Post hoc two-way clustering by country and time gives p = 0.19 to 0.20. In a post hoc, descriptive rotation, the true calendar is more negative than all 235 shifted calendars at least 60 days away, itself a post hoc exclusion (rank fraction 1/236 = 0.004, not a p-value, as shift exchangeability is unestablished; 1/296 = 0.003 when only shifts under 30 days, which share Ramadan days, are excluded). The Muslim-majority minus placebo contrast, -0.9%, is not significant under any method (p = 0.09 to 0.43). Within this model the primary interval (about -4% to +0.8% per Ramadan month) cannot separate a deficit of 1-2% from zero and does not bound the behavioural effect, which non-Muslim residents and timing error may dilute and residual calendar confounding may offset.