|
International aid agencies often prefer to use a “geographic targeting” approach in disbursing funds in their poverty reduction programs. The approach allows them to allocate resources in accordance with the different poverty levels of different
geographic areas.
Traditional household socio-economic surveys do not usually satisfy such a requirement because they are representative only at a
highly aggregated level. However, this challenge has been partially met with the recent developments in “poverty mapping methodology” which make it possible to derive more accurate estimates of the poverty level in small geographic areas.
A number of poverty mapping methodologies are currently available. According to Assistant Professor Tomoki Fujii from the School
of Economics and Social Sciences (SESS), if all the methodologies lead to an identical outcome, the choice of the methodology
would not be an issue. “One could simply choose the methodology that requires the least cost,” he said. However, as he pointed out, in practice, it is unlikely that different methodologies would lead to the same
result. Yet, there is little empirical evidence to date on how much the choice of the methodology matters.
In a recent paper, “How Well Can We Target Resources with “Quick-and-Dirty” Data: Empirical Results from Cambodia” (Tomoki Fujii, 2006, SMU Economics Working Paper and in submission to World Development), Prof Fujii tried to provide an evaluation of the effects of using different poverty mapping methodologies. Specifically, he compared the poverty rankings obtained by two of the most commonly used methodologies, “small-area estimation” and “principal component score”. The former derives the poverty mapping by combining a census and survey data set through a regression model while the latter makes use of various indicators from the commune-level data. The commune-level data
are considered quick and dirty as they are easy to collect, and the quality of the data may not be substantially worse than household-level data.
Prof Fujii found that the poverty rankings from the two methodologies were in most cases positively correlated. He also checked the two rankings against the commune classification data base (which contains the
subjective ranking of poverty of the communes in the same district) and explored the different effects by ecozones with the help of commune classification database.
He was able to offer plausible explanation of discrepancies observed such as the different price system in rural areas. He found that the commune classification database provides a ranking closer to the principal component score ranking than the small-area estimation ranking.
In the paper, Prof Fujii also evaluated the loss of efficiency when researchers use a “quick-and-dirty” poverty mapping method. Using the “small-area estimation” method as a benchmark, he analyzed the different efficiency levels of the various methods. He concluded that the efficiency losses depend to a large extent on the size of the poverty reduction budget (and hence how ambitious the poverty reduction program is). For example, the principal component score could capture one-third to one-fourth of the potential gains from commune-level targeting depending on the size of the budget. How significant such a result is depends on whether additional data collection is required to carry out small-area estimation, as well as the cost of additional data collection. “This is the first time that the magnitude of such effects has been estimated in the literature,” said Prof Fujii.
Prof Fujii found that when the budget for the poverty reduction program is limited, the difference in efficiency gains arising from using different poverty mapping methods may not be big. It may make sense in this case to use the “quick-and-dirty” method. On the other hand, when the budget is sufficiently large, the budgetary gains may be large enough to justify using the more elaborate “small-area estimation” method.
|
