COMPSTAT 2016: Start Registration
View Submission - COMPSTAT
Title: Poverty mapping in small areas under a two-fold nested error regression model Authors:  Isabel Molina - Universidad Carlos III de Madrid (Spain) [presenting]
Domingo Morales - University Miguel Hernandez of Elche (Spain)
Yolanda Marhuenda - Universidad Miguel Hernandez de Elche (Spain)
Abstract: When the target population is naturally divided in subpopulations at two nested aggregation levels (e.g. in provinces and counties within provinces), or when the sampling design has two stages as is usual in many household surveys, it is reasonable to assume a two-fold nested error model including random effects at the two levels of aggregation, domains and subdomains. A previous empirical best method for poverty mapping is extended to a two-fold model of this kind, when the target parameters are separable. We provide analytical expressions for the EB estimators of poverty incidences and gaps obtained under the two-fold model and also a Monte Carlo algorithm for approximation of EB estimators of more complex domain or subdomain parameters. The obtained EB estimates of subdomain parameters have the good property of being consistent with the corresponding domain estimate. We provide a bootstrap estimator of the mean squared error (MSE) of EB estimators. In simulations, we compare the EB estimators of poverty incidence and poverty gap obtained under the two-fold model with the EB estimators obtained by considering a model with only domain effects or only subdomain effects, when all subdomains are sampled or when there are unsampled subdomains. Results are applied to poverty mapping in counties of the Spanish region of Valencia by gender.