Small area estimation of poverty proportions under unit-level temporal binomial-logit mixed models. (2018) Spriger Link. 27 (2) : 270-294
Hobza, T. (Department of Mathematics, Czech Technical University in Prague); Morales, D. (Operations Research Center, Miguel Hernández University of Elche); Santamaría, L. (Operations Research Center, Miguel Hernández University of Elche)
Abstract. Poverty proportions are averages of dichotomic variables that can be explained by unit-level binomial-logit mixed models. The change between the poverty proportions of two consecutive years is an indicator describing the evolution of poverty. This paper applies a unit-level temporal binomial-logit mixed model for estimating poverty proportions and their changes. The model parameters are estimated by the maximum likelihood method for the Laplace approximation of the loglikelihood. The empirical best predictors (EBP) of proportions and changes are calculated and compared with plug-in estimators. The mean squared error of the EBP is estimated by a parametric bootstrap. A simulation experiment is carried out to study the empirical behavior of the EBP and the plug-in estimators. An application to the estimation of poverty proportions and changes in counties of the region of Valencia, Spain, is given.