Sources: Child mortality
This page describes how Gapminder has combined data from multiple sources into one long coherent dataset with Child mortality under age 5, for all countries for all years between 1800 to 2100.
Documentation – version 10
Download » Excel file with data for countries, regions and global total– v10
Summary documentation of v10
— 1800 to 1950: Gapminder v7 ( In some cases this is also used for years after 1950, see below.) This was compiled and documented by Mattias Lindgren from many sources, but mainly based on www.mortality.org and the series of books called International Historical Statistics by Brian R Mitchell, which often have historic estimates of Infant mortality rate which were converted to Child mortality through regression. See detailed documentation of v7 below.
— 1950 to 2016: UNIGME, is a data collaboration project between UNICEF, WHO, UN Population Division and the World Bank. They released new estimates of child mortality for countries and a global estimate on October 17, 2017, which is available at www.childmortality.org. In this dataset almost all countries have estimates between 1970 and 2016, while roughly half the countries also reach back to 1950.
— 1950 to 2100: UN WPP, World Population Prospects 2017 provides annual data for Child mortality rate for all countries in the interpolated demographic indicators, called WPP2017_INT_F01_ANNUAL_DEMOGRAPHIC_INDICATORS.xlsx, accessed on September 2, 2017.
How the sources were combined
In general we connected our historic estimates from Gapminder v7 to the earliest available year with data in UNIGME or if it didn’t have data, we used UN WPP form 1950 and on, until UNIGME had data. Which ever of these two sources has data first we smoothly transition our historic trend over a period of 20 years from the number in our historic sources to the reach the level of the first datapoint in UNIGME or UN POP. This means that in the period 1940 to 1980, depending on data availability, different countries are moving between sources at different point in time. In most cases, the sources are very similar for the overlapping years.
That describes the principle followed for most countries, but then we made exceptions in 47 cases where we preferred Gapminder’s handmade series instead of UN POP, for the period 1950 to the start of UNIGME data, because UN POP has smoothed their trends which makes events in single years (like genocides) not show up as dips or peaks in single years, instead they are smoothly decreasing or increasing the trend over a decade. Also sometimes, the UN Pop number is far away from UNIGME while Gapminder’s historic estimates tend to be closer.
Global and regional averages
Our global trend for Child mortality rate is using the UN IGME data for the period 1970 to 2016. All other years is a weighted mean of countries data. The proper way to calculate the global child mortality, would require estimating the total number of child births and child deaths each year. But we don’t have good estimates of the number of births, so instead we have used a proxy: Fertility rate multiplied by population. This method gets us very close to the properly calculated UNIGME numbers. For 1990, UNIGME has 93.4, and our weighted average is 96.6. We have linked our weighted average for the world, to the UNIGME series, by using the rate of change before 1990, and apply that backwards in time, so the whole series is moved down to meet UNIGME in 1990.
This graphs shows the four lines. The purple is our weighted average that is slightly higher than UNIGME in 1990 (the green line). And the black dotted line is Gapminder v10 after the whole line is adjusted downwards to hit UNIGME in 1990. The blue line shows UN WPP which have much fewer peaks than our trend, because detailed yearly events are smoothed out over decades.
The regional averages were calculated with the same weight method.
In future version of our trend for Under five mortality we hope to include lots of more temporary peaks, by taking the guesstimated dips in life expectancy and estimate how much child mortality must have increased during times of catastrophes.
Detailed documentation and feedback
For transparency we provide the files we used to calculate this data here, but we haven’t had time to clean them up and document all the steps in the process, see our data method overview. Any questions about the data and suggestions for how to improve it are always very welcome: in our data forum.
These are Gapminder’s historic estimates compiled by Mattias Lindgren.