What Can Satellite Night-Time Photos Tell Us About the State of Global Poverty?

Economists Maxim Pinkovskiy of
the Federal Reserve Bank of New York and Xavier Sala-i-Martin of
Columbia University have
written a paper on a new way to assess the state of global
poverty.

Over at the Adam Smith Institute’s
blog Tim Worstall summarizes the problem Pinkovskiy and
Sala-i-Martin are addressing:

A government that hasn’t quite worked out how to collect the
rubbish isn’t going to be all that good at counting the people nor
what they do or earn.

The upshot of this is that we really don’t know how many poor
people there are out there. We know there are fewer than there used
to be, and the poor are a very much smaller portion of the growing
population than used to be true, but we really only know trends
rather than actual numbers. Basically because the numbers we’ve got
for things like GDP and inequality are so sketchy themselves.

What Pinkovskiy and Sala-i-Martin are proposing is the
following:

In new research we suggest such a data-driven way to assess the
relative quality of national accounts and survey means by using a
third, independently collected data source on economic activity.
This data is satellite-recorded luminosity at night as measured by
weather satellites of the National Oceanic and Atmospheric
Administration (NOAA). The advantage of this data is that
measurement error in night-time lights is unrelated to measurement
errors in national accounts or surveys, making lights an
independent measurement of true income, against which we can assess
the other measurements.

To disentangle whether national accounts or survey means are
closer to true income we postulate a factor model in which national
accounts GDP per capita, survey means and night-time lights are
linear functions of unobserved, true income (and, potentially,
other covariates), perturbed by random disturbances.

What is particularly interesting about Pinkovskiy and
Sala-i-Martin’s paper is that it concludes with some encouraging
news:

Under our procedure, developing world poverty declines from
11.8% in 1992 to 6.1% in 2005 and 4.5% in 2010, much lower than the
path constructed by giving a weight of 1 to the surveys, which
entails poverty falling from 42% to 20.5% between 1992 and 2010. We
run a battery of robustness checks on our findings; under the ones
most favourable to replicating the survey-based results, the
largest that we find developing world poverty to be in 2010 is
12%.

Read more from Reason.com on poverty here.