Small Domain (Area) Estimation: An appraisal and experiences from applications in National Statistical Institutes in Europe

Nikos Tzavidis, Universidade de Southampton, Inglaterra

Abstract:

In the last decade there have been a number of signi_cant methodological developments in the small area estimation literature. Recent methodologi cal advances include spatial and temporal, outlier robust, non-parametric and semiparametric extensions of the unit and area level small area models,models for discrete outcomes, methods for estimating small area distribution functions, new approaches to MSE estimation, estimation under informative sampling and accounting for measurement errors. At the same time, applied work has focused on the implementation of some of the cutting edge small area methodologies in the production line of National Statistical Institutes (NSIs).

In this talk I will start by presenting an appraisal of some of the newly proposed methods for small area estimation. I will then attempt to link the world of academic research to the real world of producing national statistics.

I will do so by presenting recent experiences from applying small area estimation techniques in NSIs in Europe. In particular, I will describe aspects of the current SAE programme of the UK O_ce for National Statistics focusing mainly on the estimation of income deprivation and income distributions. In addition, I will describe the application of small domain estimation methodologies to business survey data from the Central Bureau of Statistics in the Netherlands and speci_cally on the use of outlier robust methods. The talk will be concluded with some reections and proposed directions for future work.