The main empirical results of this paper are: 1) Global poverty rates decline
between 1970 and 2006. This is true for poverty lines ranging from $1/day to $10/day. 2)
Global poverty counts decline between 1970 and 2006 for poverty lines from $1/day to
$3/day. The total number of poor people has declined by more than 617 million if we use
the $1/day line and by more than 780 million if we use the $2/day line. For higher
poverty lines, poverty counts increased during the early years but are all declining by
2006. 3) Global income inequality has fallen between 1970 and 2006. This is true for the
Gini coefficient, for a wide variety of Atkinson indexes and General Entropy indexes as
well as the 90th-to-10th and the 75th-to-25th percentile ratios. 4) We systematically analyze
the normative effects of changes in the world distribution of income using a strongly
microfounded definition of welfare as the certainty equivalent of a lottery over all
incomes in the world. We find that world welfare increases at increasing rates during our
sample period, and records dramatic increases whether computed across citizens or across
countries, notwithstanding increases in unweighted between-country inequality. Total
growth in world welfare measured is estimated to be between 77% and 160%, with most
estimates over 100%. 5) At the regional level we observe that poverty rates and GDP per
capita behave as “mirror images” of one another: whenever GDP grows, poverty tends to
decline and whenever poverty declines, GDP tends not grow. 6) Poverty has declined
substantially in East and South Asia, and has recently began declining in Africa.
We show that our conclusions are robust to a general sensitivity analysis, and in
particular, to three key areas of uncertainty: 1) uncertainty over the functional form of the
country income distribution, 2) uncertainty over potential nonresponse biases in
household surveys, and 3) uncertainty over the correct method of computing a PPPadjusted
GDP series. We show that the robustness extends not only to the global trends
of falling poverty and inequality, but also that finer trends persist under specification
changes. In particular we robustly demonstrate that, 1) poverty exhibits a “tsunami”
effect, in which poverty declines decelerate for lower poverty lines and accelerate at
higher ones, 2) poverty becomes an essentially African phenomenon, and 3) most of the
decline in inequality is a decline in population-weighted between-country inequality.
Our data allows us to give a progress report on the first Millennium Development
Goal of halving poverty from 1990 to 2015. Table 12 shows that estimates from our
modifications in the sensitivity analysis indicate that so far, poverty has fallen by about
30% in the 16 years since 1990, giving the world ample time to reduce poverty by a
further 20% of 1990 levels. If we accept the World Bank’s recent PPP revision, then
poverty has fallen by about 58%, and the first MDG has been achieved, since the PPP
revision assigns to the 1990s a large part of China’s poverty decline that the Penn World
Tables and Maddison assign to the 1980s. Importantly, a large part of the decline in
poverty has taken place in Africa. Using the Penn World Tables as our source of GDP,
we see that Africa has decreased poverty by 20-25% from 1990 levels, making it likely
that it will come close to, or perhaps even achieve the MDG within its continent.