Page 168 - Contributed Paper Session (CPS) - Volume 8

CPS2227 Jonathan Haughton
Even when survey data are of good quality, the steps required to arrive at
reliable and valid measures of poverty are sufficiently intricate that researchers
may draw very different conclusions from the same data. A dramatic
illustration of this is found in the case of Rwanda, whose GDP has grown by
xxx% per year over the past decade. According to the National Institute of
Statistics of Rwanda (NISR), the proportion of Rwandans in poverty fell from
45% in 2011 to 39% in 2014 and 38% in 2017, although the drop over the
latter period was not statistically significant (NISR 2018). A recent study that
appeared in the Review of African Political Economy used the same survey
data to conclude that the poverty rate in Rwanda rose from 52% in 2014 to
58% in 2017 and was higher in 2017 than in 2001 (ROAPC 2019).
This provides support for the contention of Pogge and Wisor (2016 pp. 4-
5) that “The result of various internal challenges … the methods used for
setting poverty lines and calculating individual achievements against this
standard … is that the scope, distribution, and trend of poverty within and
across countries varies greatly depending on which assumptions are used.” If
poverty reduction were closely linked to GDP growth – with a stable income
elasticity of poverty – then it would be practicable to predict poverty rates
based on anticipated economic growth as done by Chandy et al. (2013), and
this may be reasonable over long intervals (Dollar and Kraay 2001), but the
relationship is not close enough in the short-run for this to be compelling. We
thus need to rely on survey data to track the evolution of poverty over time.
In this paper we examine the sensitivity of measures of poverty to the
choices made by analysts about these “internal” decisions, mainly using survey
data from Rwanda, but with appropriate reference to the experience of other
countries. We focus on four (of the many possible) “internal” issues where the
assumptions made by the analyst may matter: valuing auto consumption,
adjusting for prices over time and space, specifying adult equivalents, and
establishing a poverty line. Of these, the most difficult is getting the prices
right.
In what follows we set out each issue briefly and indicate our preliminary
findings.

2. Valuing autoconsumption
In poor countries, a significant amount of household consumption comes
from home production. Since this autoconsumption is not sold, the question
arises of how best to value it. Many surveys ask households how much they
could get if they were to sell the good or service in the marketplace. A
potential difficulty here is that there is some evidence that households tend to
understate the selling price. Also, it may not be ideal to value (say) sweet
potatoes at the buying price for some households (if they buy the good), and

157 | I S I W S C 2 0 1 9