|
Conference Paper |
|
| Targeting the poor in Egypt, a ROC approach |
|
Magued Osman, Enas Zakareya and Walaa Mahrous
Abstract
We adopt an inexpensive framework that can be easily implemented for targeting the poor in Egypt, using the Social Contract Survey-Egypt 2005 database. Disaggregate micro-level household expenditure/income data are not available to all Egyptian
researchers. This represents the main challenge that we have faced in attempting to formulate an antipoverty targeting program for Egypt. Hence, we employ a model that does not depend on these data. We use the principal component factor analysis
(PCFA) technique to build a proxy index for household wealth using various asset ownership and socio-demographic indicator variables. Considerable attention is devoted to the validation of the coherence of the computed index and to the examination of its plausibility as an instrument for characterizing poverty. The wealth
index is used to identify the poor relative to the non-poor households. We refer to the households with values on the wealth index that belong to the bottom 20% of the population as poor. Using this classification together with an optimal set of poverty
covariates, which are easy to get data on in Egypt, we derive composite poverty indicators (CPIs) for poverty monitoring and targeting. The analysis establishes the importance of the level of education of the household head and to a lesser extent the
housing quality and the place of residence as key control indicators of household poverty that enter into the composition of the CPIs. The Receiver Operating Characteristic (ROC) curves are utilized to examine the efficiency of the estimated CPIs, to characterize the tradeoff between the identification of the poor and the exclusion of the non-poor and to pinpoint pragmatic guidelines for poverty alleviation for Egypt as a whole and across the urban-rural divide. The robustness of the CPIs and of the ROC analysis is examined by testing the effect of targeting the bottom 40% rather than the poor 20% of the population. We find that stylized programs, which are specifically designed to target the poor, are relatively more efficient.
Published: 2006
Number: 132006017
Type: Conference Paper
|
