Data: Extreme Poverty Rate — v1

Updated: April 9, 2018

Before 1980 we have combined data form three sources: Gapminder Income Mountains (GM), Bourguignon and Morrisson (BM), Our World In Data (OWID), and van Zanden[1] (VZ), as described in the sheet "Comparing sources".

Gapminder has estimated the extrem epoverty rate of the world since 1800 to 2040, by aligning trends fomr multiple sources.
— For the period 1980 to 2015, we use the World Bank estimates we no modifications.
— Before 1980, two historic sources have influenced the line.
——Bourguignon and Morrisson (2002) , uses naitonal accounts to estimate extreme poverty rate, and presents to different rates wihth slightly different treasholds: "‘these definitions correspond to poverty lines equal to consumption per capita of $2 and $1 a day, expressed in 1985 PPP.’" This source gives higher rates back in 1820 ( 94.4% and 83.9% respectivly), compared to the next source. (The sources uses different benshmark years for their constant year for price adjusted dollars. But that difference is not sufficient to explain the huge variation.
—— van Zanden et.al. 2014: is the second historic source, using estimates of historic inequality in countries, covering 75% of humanity, based on historic records, like length of soldiers in military archives. They generate distribution curves of number of people over income, and then extract the share of poeple below two treasholds, published in a working paper: "World Income Inequality 1820-2001". Their dollars are expressed in price levels of 1990. Their estimates for extreme poverty rate in 1820, are , $1/day : 39%, $/day: 73%.

The Gapminder estimates ends up between these two. We have generated income distibutins shapes covering not only 75% of humnanity, but including all countries, available here: gapm.io/incm. These curves uses PPP 2011, like the recent numbers form World Bank. These curves are based on other historic estimates by Gapminder for the three indicators:: Mean houshold per cap,
income, Gini and Population, as documented here: gapm.io/doc_incm.

We draw curves for all countries all years showing the amount of people on different incomes, which we call Income Mountians. From these distribution curves we extract the share of people below a given income, with the same method used by van Zanden. This gives us a complete time-series for extreme poverty worldwide, for every year: estimating the share of mankind below $1.9/day (in intenrational dollars PPP2011 which is the same unit as the official World Bank esitmates, and our number ends up being . With our historic income eand gini sitmates we have esitmated the extreme poverty rate back to 1800, and our figures end up between the estimates from the two sources above. We think van Zanden et.al. is most likely underestimating extreme poverty (as the historic data is leaving out 25% of humanity, mostly from poorer countries). To avoid relying solaly on our numeric exersice, we use the average of our

Sources as seen used in the sheet "Comparing sources"

See: https://ourworldindata.org/grapher/world-population-in-extreme-poverty-absolute?stackMode=relative&country=OWID_WRL

The average of bm1 & bm2 meets the wb level very well both in year , around 44% in 1980, and 35% in 1992

Max Roser picked the higher line from BM and blend it to WB at the end

https://ourworldindata.org/grapher/world-population-in-extreme-poverty-absolute?stackMode=relative&country=OWID_WRL

None of these trends are close to the world bank esitmates for the years