In response to high demands for lower level poverty estimates, the National Statistical Coordination Board releases provincial estimates, in addition to the national and regional, starting with the 1997 FIES. However, estimates of the coefficients of variation (CV) of several provincial estimates indicate that the resulting poverty measures are not reliable. Making a decision based on unreliable poverty statistics is very risky especially if the decision to be made relates to the welfare of poor families. Such unreliable poverty statistics may also lead to incorrect targeting of the right beneficiaries of the poverty alleviation program. Hence, this paper provides alternative ways of coming up with subnational statistics (i.e., provincial and municipal/city-level data) that yield lower CVs than those of the official ones. This refers to the small area estimation (SAE) technique, a model-based approach to produce provincial or even municipal-level data. With a good predicting model, the SAE technique has a lot of potential in providing reliable subnational estimates for poverty reduction efforts.
poverty statistics small area estimates (SAE) synthetic estimation composite estimation regression-synthetic empirical best linear unbiased prediction estimator (EBLUP) Elbers Lanjouw and Lanjouw (ELL) estimation procedure