Page 169 - Contributed Paper Session (CPS) - Volume 8

CPS2227 Jonathan Haughton
at a different price (the selling price, if they consume their own production) for
others.
One approach would be to value autoconsumption using some cluster-
level median household- reported price, which should reduce idiosyncratic
variation in prices, but at the risk of overlooking variations in quality. Another
possibility would be to collect prices in a separate survey, with the express
purpose of using them to value household consumption. And a third
possibility, used in Rwanda, is to use prices collected by the NISR for
constructing the consumer price index.
Using data from 2017, we found that there is no systematic difference
between the median prices reported by households, and the prices collected
in the clusters where the households reside, although the household-reported
prices have a very wide variance.

3. Adjusting for prices over time and space
The household survey data needed to measure poverty are collected in
different areas of the country at different times of the year. There is
geographical and temporal variation in prices, and the value of household
spending needs to be deflated in order to take this variation into account.
Some, but certainly not all, household surveys collect information on both
values and quantities of items consumed, and so one can infer unit costs,
which are similar to prices. In principle these could be used to deflate
household expenditure, although in practice they may not be very accurate,
given variations in the quality of products. In a few cases, a complementary
survey collects price data from the areas in which households are surveyed,
but the quality of these data often suffers, since the enumerators may not be
skilled at correcting for differences in quality or type. Sometimes the price data
collected for the consumer price index (CPI) are used, although in some
countries these numbers are only available for urban areas, which makes them
poor guides to the prices faced by poor, largely rural, households.
Given a set of prices, it is possible to deflate consumption for each
household, as done by Kenya (2007) and recommended by Deaton and Zaidi
(2002). Since these indexes in effect use the consumption weights in the end
period, they are Paasche indexes, and tend to understate inflation.
It appears to be more common to create a “poverty price index” that uses
commodity weights that reflect the experience of households close to the
poverty line – perhaps the poorest (say) 40% of the population (as in Rwanda
2018), or those whose consumption is expected to be close to the poverty line
(e.g. Vietnam). More often than not these indexes use the end weights, and so
are also Paasche indexes, although if there are panel data, more satisfactory
indexes may be applied (see Haughton and Khandker 2009).

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