RIVISTA INTERNAZIONALE DI SCIENZE SOCIALI - 2008 - 4

Methods for small area estimation have received much attention in recent years due to growing demand for reliable small area statistics that are needed in formulating policies and programs, allocation of government funds, making business decisions and so on. Traditional area-specific direct estimation methods are not suitable in the small area context because of small (or even zero) area-specific sample sizes. As a result, indirect estimation methods that borrow information across related areas through implicit or explicit linking models and auxiliary information, such as census data and administrative records, are needed. This paper provides an introduction to small area estimation with emphasis on explicit model-based estimation. Methods covered include «off-the-shelf» re-weighting methods, simulated census methods used by the World Bank and formal empirical Bayes and hierarchical Bayes methods, based on explicit models. Formal model-based methods permit the estimation of mean squared prediction error and the construction of confidence intervals.

Demands of regional statistics combined with pressure to reduce costs and response burden have lead to great interest in Small Area Estimation. Both researchers and practitioners take part in a largely very successful development that still is moving rapidly. However, the National Statistical Institutes in Europe have been rather hesitant to implement SAE, partly because of the different tradition of SAE in terms of statistical inference. NSIs are obliged to as far as possible publish officials statistics that are based on estimators with negligible bias. Fear of model misspecification has been a hindrance to wide application of SAE. Use of a model is now seen as a quality issue. Communication of methods and the resulting quality of statistics is an issue that NSIs have recently given specific attention to.

Small area estimation techniques are becoming more and more important for the production of official data. Since 2002 ISTAT has applied small area estimation methods to disseminate estimates of employment and unemployment rates for local labour market areas; furthermore several experimental studies have been carried out for the estimation of the poverty rate at province level. In this paper details of the main studies carried out regarding the application of small area techniques to ISTAT household surveys are reported.

One of the contexts where small area estimation techniques have proved their potential is the analysis of data collected in national labour force surveys to obtain estimates for small geographical domains. Applications of small area estimation methods to data from labour force surveys have recently been considered in Italy. This paper gives a review of specific problems, data and opportunities for the application of small area estimation models for producing reliable information at provincial and sub-provincial level in Italy on labour force aggregates. Some new developments stimulated by the application of small area estimation models to the analysis of labour force survey data are also discussed.

This paper examines the performance small area of welfare estimation. The method combines census and survey data to produce spatially disaggregated poverty and inequality estimates. To test the method, predicted welfare indicators for a set of target populations are compared with their true values. The target populations are constructed using actual data from a census of households in a set of rural Mexican communities. Estimates are examined along three criteria: accuracy of confidence intervals, bias and correlation with true values. We find that while point estimates are very stable, the precision of the estimates varies with alternative simulation methods. Precision of estimates is shown to diminish markedly if unobserved location effects at the village level are not well captured in underlying consumption models. With well specified models there is only slight evidence of bias, but we show that bias increases if underlying models fail to capture latent location effects.