CollinsAnyona/african-child-survival

Which African countries beat the child-mortality trend and why? UN IGME/World Bank panel analysis: SDG 3.2 pace, beat-the-trend residuals, and an honest ML-vs-baseline comparison. # The Child Survival Revolution, Unevenly Shared An end-to-end data analysis of under-5 mortality in Africa since 1990 — who beat the trend, who fell behind, and whether the SDG 3.2 target (under-5 mortality ≤ 25 per 1,000 live births by 2030) is still reachable — using UN IGME estimates and covariates from the World Bank Open Data API, 54 countries, 1990–2024. **Why it matters:** a child born in the median African country in 1990 faced roughly a 1-in-7 chance of dying before age five; today it is closer to 1-in-20. That is one of the largest improvements in human welfare on record — and it is wildly uneven. Knowing *which* countries outran their starting conditions, and what observable factors travel with faster progress, is the difference between celebrating an average and understanding a continent. ## Headline findings **1. The revolution is real.** 43 of 54 African countries at least halved under-5 mortality between 1990 and 2024, and 26 cut it by two-thirds or more. The median African country went from 152 to 49 deaths per 1,000 live births (an annualised decline of 3.0%/yr); the Sub-Saharan aggregate fell from 178 to 71 while the world went from 94 to 37. **2. But SDG 3.2 is effectively out of reach for most of the continent.** Only 9 of 54 countries are on track for the 2030 target — and every one of the nine (Egypt, Algeria, Mauritius, Seychelles, Morocco, Tunisia, São Tomé & Príncipe, Libya, Cabo Verde) has *already* met it. Not a single country still above 25 is projected to get there by 2030 at its own recent pace: the median off-track country needs to cut mortality by 13.5%/yr but is managing 3.2%/yr. For context, 135 of 196 countries worldwide are on track — Africa's 9/54 is the global exception, not the rule. **3. Starting point is not destiny.** Regressing each country's decline pace on its (log) 1990 level, the starting point explains only 11% of the variance in progress (slope p = 0.015). Ranking countries by the residual — how much faste …