Old news, perhaps, but well worth reading and rereading. Deaton discusses the recent revaluation of the global poverty to show why there was such a large change in the number of poor people. The primary point that has been emphasized from the official story has been that, while the developing world is poorer (relative to developed countries) than previously imagined, the downward trend has remained over time. The principle change they point to is the revision of PPP price data for China.
Deaton points out that this is not the most important change. The most important change was removing India from the set of countries whose poverty lines are averaged to form the world poverty line. This had the effect of raising the real poverty line. Inasmuch as there are nearly 200 million Indians between the old line and the new line, this "Indianization" or "reIndianization" of poverty changes how we frame the aid and development debate. It changes the MDGs. "In this particular example ... a small increase in Indian incomes caus[ed] a large increase in Indian and other countries' poverty counts. In effect, India and the world have become poorer because India has become richer!"
And changing the poverty line changes inequality. Because we changed how poor we think the poor countries are, it made it look like the rich countries are richer than they used to (from the perspective of the poorest countries).
Deaton complains overall that each major revision to the PPP series throws serious and important changes into how we perceive the world. Before 1993, it was accepted wisdom that Asia was poorer than Africa. Afterward, Africa took center stage. "It seems impossible to make statements about changes in world poverty when the ground underneath one's feet is changing in this way," he wrote at the time.
He enters into a detailed and enlightening discussion of how PPP numbers are created. His primary criticisms center around how housing is imputed (when national governments don't impute it, it throws the poverty numbers off; the method was designed to make GDP numbers comparable but are unfortunate for poverty), which alone can reduce the 2005 poverty count by more than 100 million people.
In addition to fixing the housing imputation problem, he advocates either picking an absolute poverty line (India, or poorest 15 countries, or $1/day, whatever you like) and sticking with it or using Gallup-esque surveys of how people evaluate their own life and comfort.