Abstract

This study titled as “A Look at EBLUP and other Small Area Estimates of Poverty Incidence at the City/Municial Level” is an attempt to respond to the growing demand for more reliable small area statistics. It focuses on two approaches to small area estimation: synthetic estimation and empirical best linear unbiased predictor (EBLUP). Estimates of poverty incidence and measures of the reliability are obtained for the cities/municipalities of Nueva Ecija and Pangasinan, the two provinces chosen for the study. Because the usual sample survey estimators or direct estimators typically do not yield reliable estimates for small areas or subdomains, alternative procedures that “borrow” information from similar small domains so as to increase the accuracy of the resulting estimates were resorted to. This technique of pooling information of small domains has been referred to as synthetic estimation. The synthetic estimator may be viewed as an extension of the usual ratio estimator to g groups (the population is partitioned into large domains indexed by g). Synthetic estimates of poverty incidence for Nueva Ecija ranged from 15.6 (Cabanatuan City) to 42.7 (San Isidro), and those for Pangasinan ranged from 44.1 (Sta. Barbara) to 49.9 (Bani). As for the weighted least square (WLS) and EBLUP estimators, the EBLUP appears to be more precise. Its implementation is constrained, at present, by the need to develop the programs for diagnostic checking. In both cases however, adequacy of the model depends in large measure on the available explanatory variables. A low R2 for the fixed regression part of the model does not inspire much confidence in the estimates. EBLUP estimates for Nueva Ecija ranged from 21.6 (Cabanatuan City) to 54.1 (Lupao); WLS estimates for the same province ranged from 20.0 (Cabanatuan City) to 52.1 (Gen. Tinio). EBLUP estimates for Pangasinan ranged from 17.4 (Sto. Tomas) to 73.4 (Infanta); WLS for Pangasinan ranged from 13.4 (Sto. Tomas) to 66.8 (Manaoag).