Abstract

The growing need for disaggregated statistics for specific geographic areas and small groups of the population has gained increased attention in National Statistical Offices in Latin America. Small Area Estimation (SAE) procedures methods can adequately fulfil these needs; these techniques allow for achieving accuracy and precision beyond the limits imposed by household survey samples, and direct estimators. Many countries in Latin America have begun to use SAE techniques to provide official statistics on several topics, for example, unemployment and poverty.

This session, organized by the Division of Statistics of UN-ECLAC, will present some regional advances in the use of SAE methods to obtain disaggregated estimates through the integration of different data sources (surveys, administrative records censuses, and satellite imagery). The session will have presentations on methodologies, tools, and national experiences in producing disaggregated statistics using SAE techniques and transforming SAE from experiment to official data production. The session will cover the next four talks:

1. The PNAD Contínua is the largest household survey in Brazil and the source of the main sociodemographic and labour force indicators. Even collecting data from more than 600,000 people per quarter, the sample size is not always sufficient to disclose figures, such as indicators related to SDG 4, in domains needed by the policymakers. Thus, the main objective of this study is to provide an SAE for educational indicators, using PNAD Contínua data, in order to fill that gap.

2. Peru is divided into ~2000 districts, and each one of them has a local authority conducting social policy. Thus SAE turns out to be an incredibly helpful tool for evidence-based policymaking. In this talk, we will present how SAE and Machine Learning techniques have been used to derive novel indicators related to pressing problems in the country: vulnerability to poverty and children's anaemia.

3. Quarterly unemployment estimates for the Brazilian Labour Force Survey using state-space models in small areas. The study aims to produce quarterly estimates of unemployment for small areas inside the Brazilian states based on time series models for repeated surveys. This application allows the production of seasonally adjusted series and difference indicators taking to account the sampling error.

4. Based on the new Multiple deprivation index for Latin American countries produced by the Economic Commission for Latin America and the Caribbean, this talk shows the case study of Colombia to obtain small area estimates. This country counts with a recent population census providing most of the information required to compute the multiple deprivation index in small domains. However, for two indicators the information at the unit level is not available, for which different small area estimation methods are implemented. A parametric bootstrap algorithm is used to provide uncertainty measures.