Abstract

Obtaining reliable estimates in small areas is a challenge because of the coverage and periodicity of data collection. Several techniques of small area estimation have been proposed to produce quality measures in small areas, but few of them are focused on updating these estimates. By combining the attributes of the most recent versions of the structure-preserving estimation methods, this paper proposes a new alternative to estimate and update cross-classified counts for small domains, when the variable of interest is not available in the census. The proposed methodology is used to obtain and update estimates of the incidence of poverty in 81 Costa Rican cantons for six postcensal years (2012 - 2017). As uncertainty measures, mean squared errors are estimated via parametric bootstrap, and the adequacy of the proposed method is assessed with a design-based simulation.