Poverty mapping is one of most common applications in small area estimation. Many examples are available for the proportion of population living below the international or national poverty line (indicators 1.1.1 and 1.2.1).

World Bank applications

The World Bank proposed a poverty mapping process that was conducted in several countries. Based on surveys and additional data sources, various poverty and inequality estimates such as the Foster-Greer-Thorbecke poverty estimates and the Gini coefficient were derived.

The report More than a pretty picture - Using poverty maps to design better policies and interventions published in 2007 shows case studies for the countries Albania, Bolivia, Bulgaria, Cambodia, Yunnan Province (China), Ecuador, Indonesia, Mexico, Morocco, Sri Lanka, Thailand and Vietnam that describe all poverty mapping steps and also lessons learned. Hence this can be a good starting point for a new poverty mapping study. 

In 2005, the World Bank provided technical assistance to the Philippine national statistical system to leverage on small area estimation techniques to produce municipality- and city-level poverty statistics. The Philippine Statistics Authority conducts the Family Income and Expenditure Survey (FIES), which is the main source of official poverty statistics in the country, every three years. The small area estimation technique used in the Philippines is based on the ELL method. It entails regressing (log) per capita household income from the FIES with auxiliary information from the FIES, the Labor Force Survey, and the Census of Population and Housing. The model regressors include survey-obtainable variables such as educational attainment of the household head and other household characteristics, and census-derivable information like average family size in a village, and other village-level information. Since small area poverty statistics became available in 2005, numerous government agencies have used these data as inputs for formulating and implementing poverty reduction programs.  For example, the Philippine Department of Social Welfare and Development (DSWD) used the estimates to identify poor municipalities for its National Household Targeting System for Poverty Reduction (NHTS-PR) data collection.

Poverty estimation in Chilean comunas

To improve fund allocations among comunas, the Chilean Ministerio de Desarrollo Social (in the following the ministry) is required to provide poverty estimates for all 345 comunas in Chile which is the smallest territorial entity. After the evaluation of various options, the ministry decided to combine the National Socioeconomic Characterization Survey (CASEN), which is Chile's official data source for poverty statistics, with relevant administrative records. Since 2011, model-based poverty statistics are obtained for Chilean comunas.

Small Area Income and Poverty Estimates (SAIPE) program by the U.S. Census Bureau

The SAIPE program produces small area estimates of income and poverty statistics for all school districts, counties, and states. The estimates are based on several data sources such as the American Community Survey and Federal Income Tax Returns. The produced indicators do not exactly follow the definition of the SDGs but the example is added since the SAIPE program is continuously improving their approach and the disaggregation is not only spatial but also by age groups.

Mean income in Middle-layer Super Output Areas in England and Wales

Super Output Areas (SOAs) are a geographic hierarchy introduced for the reporting of small area estimates. The mean population of Middle-layer SOAs (MSOAs) ranges from a minimum of 5,000 to 7,200. To obtain different income estimates (e.g., equivalised and unequivalised) for the MSOAs, data from the Family Resources Data is combined with additional data including Census information, energy consumption and house price statistics using a linear mixed model.

The case study is explained on the homepage with more methodological details in the technical report.

Estimation of food insecurity in Malawi

FAO’s case study illustrates the implementation of a model-assisted estimation approach proposed by Kim and Rao (2012) to produce disaggregated estimates of  SDG Indicator 2.1.2, “Prevalence of moderate or severe food insecurity in the population, based on the Food Insecurity Experience Scale (FIES)”. FAO (2022) considered the case of two independent surveys conducted in Malawi in 2016: (i) the Fourth Integrated Household Survey (IHS4) and (ii) Gallup World Poll survey (GWP). IHS4 with a large sample collected only auxiliary information and GWP with much smaller sample provided information on both the variable of interest and the auxiliary variables.

The model-assisted projection method is based on a working model which results in asymptotically unbiased projection estimates. Synthetic values of SDG 2.1.2 are generated by first fitting the working model, relating the variable of interest to the auxiliary variables, to the data from GWP and then predicting the variable of interest associated with the auxiliary variables observed in IHS4. The projection estimators are obtained from GWP and associated synthetic values.

Small Area Health Insurance Estimates (SAHIE) by the U.S. Census Bureau

The SAHIE program produces small area estimates of population with and without health insurance coverage for counties, states and the cross-classification of age, sex and income categories as well as ethnicities. The estimates are based on several data sources such as the American Community Survey and Medicaid participation.

Estimation of  literacy rates in Senegal

Senegal has been devoting efforts to improve the literacy rate in the country, especially for women. Estimated literacy rates for men and women in 2017 were 64.8% and 51.9% , respectively, according to UNESCO (http://uis.unesco.org/en/country/sn ).  Areas of high illiteracy must be identified for targeting public policies. Schmid at al.(2017) developed a small area estimation model for deriving small area literacy rate estimates for the 431 communes in Senegal by gender, using mobile phone auxiliary data. The authors propose an area level model with alternative of sources data as auxiliary variables. The estimates are based on 2011 Demographic Health Survey (DHS) carried out by the National Agency of Statistics and Demography of Senegal (ANSD, Agence Nationale de Statistique et de la Démographie)  and mobile phone data covering the year 2013 (provided by the Senegalese telecommunication company Sonatel ). The small area estimates are benchmarked such that the aggregated small area estimates agree with the national estimate for the country.

Estimation of fuel poverty at different geographic levels in UK

The Office for National Statistics (ONS) provides estimates of the proportion of households in fuel poverty at local levels. Fuel poverty statistics are required  by central and local governments and may be used to target energy efficiency policies. It is determined by the household income, household energy requirements and fuel prices. Household-level data from the English Housing Survey (EHS) is combined with aggregated household-level information from the Census and Experian data by a binomial model using the logit link function that considers that each household belongs to a specific area.

The description of the case study comprises the data and methodological setup, maps with estimates and an evaluation.

Estimation of the unemployment rate in Canada

Statistics Canada uses the Fay-Herriot model to obtain the unemployment rate for 149 areas (cities) in Canada. The data used is Statistics Canada's Labour Force Survey (LFS) which is a monthly survey that is designed to produce reliable unemployment rate estimates for the 55 Employment Insurance Economic Regions (EIER).  

The case study is an application of the small area estimation system developed by Statistics Canada (Hidiroglou, 2019). The description of the case study is extensive and it gives good insights on how to evaluate the small area estimates.

Economic activity rate at commune level in Switzerland

The structural population survey is a survey covering about 200,000 people each year in Switzerland and enables the estimation of economic activity rate for groups of 15,000 inhabitants. The combination of this survey with additional data, for example old-age and survivors' insurance, makes it possible estimate the annual economic activity rate for groups of at least 100 inhabitants.

For the experimental statistics of economic activity in Swiss communes, a linear mixed model that combines the survey data with additional data is used. The executive summary gives a first overview. Extensive simulations are available in the report part 1.1. and the description of the application in the report part 1.2.

Inbound travel spending estimates at sub-provincial level in Canada

The Visitor Travel Survey (VTS) enables the estimation of statistics on the volume of international visitors to Canada and detailed characteristics of their trips such as expenditures, activities, places visited and length of stay. It can also be used to estimate inbound travel spending at sub-provincial levels. However, the direct estimates may be unreliable when the sample size is not large enough.

Statistics Canada uses an adjusted area-level model to obtain the inbound travel spending for a combination of 11 country groups and 22 tourism regions. As additional data source, Payment processors' data is included.  

A short description of the case study is available that describes but does not show the validation of the model.

Estimation of ICT indicators in Brazilian states

A department of the Brazilian Network Information Center (NIC.br) called Regional Center for Studies on the Development of the Information Society (Cetic.br) collects data about access and use of information and communication technologies (ICT) in Brazil. Data users are interested in timely publication of main ICT indicators for the 27 Brazilian states. The estimated indicators include the proportion of households with computers, and proportion of households with Internet access (which is related to 17.8.1 Proportion of individuals using the Internet). The data source is the Brazilian Annual Household Survey on ICT whose sample contains almost 33,000 households. Reliable estimates for the five larger regions North, Northeast, Southeast, South and Center-West can be produced. Direct estimation approaches by averaging estimates of consecutive years and, by pooling samples of consecutive years are employed. In addition, a single-year composite estimator considering the regions are tested for yielding synthetic estimates; and a composite estimator based on pooling samples from two consecutive years, and using the regions. The simpler approaches are chosen due to a wide range of indicators that need to be produced in a timely manner after data collection. For more information, see Bertolini Coelho et al. (2020).