Abstract

The adoption of the 2030 Agenda for Sustainable Development - with its central transformative tenet of leaving no one behind - has generated a tremendous increase in the demand of disaggregated statistics. While many of the Sustainable Development Goal (SDG) indicators are directly focused on the socio-economic inequalities that affect the most vulnerable populations, at the very core of the monitoring framework stands an overarching principle of data disaggregation stating that “SDG Indicators should be disaggregated, where relevant, by income, sex, age, race, ethnicity, migratory status, disability and geographic location, or other characteristics in accordance with the Fundamental Principles of Official Statistics” (UN General Assembly resolution 68/261).

Within this context, traditional agricultural and household surveys are the fundamental data sources to monitor many SDG indicators. In particular, six out of the twenty-one SDG indicators under the custodianship of the Food and Agriculture Organization of the United Nations (FAO) can be estimated using microdata from sample surveys of different kinds. Examples of such indicators are: 1) Indicator 2.1.2, monitoring the prevalence of moderate and severe food insecurity in the population; 2) Indicators 2.3.1 and 2.3.2, measuring respectively the average labor productivity and the agricultural income of small-scale food producers; and 3) Indicator 5.a.1, quantifying women access to ownership and secure rights over agricultural land.
This paper discusses two main limitations that traditional agricultural and household surveys present in producing granular disaggregated estimates of the four SDG Indicators mentioned above, both at the sub-national level and by disadvantaged socio-economic populations. In addition, the article presents some applications of indirect estimation techniques – such as small area estimation (SAE) –implemented by the FAO to address these key limitations.
Firstly, sample surveys are designed to produce reliable direct estimates of target parameters in planned estimation domains, i.e. disaggregation dimensions that were accounted for at the sampling design stage. On the other hand, samples of most surveys are generally not large enough to guarantee representative samples of all sub-populations, or the production of precise estimates in all possible disaggregation domains. To address this issue, the paper presents empirical applications of unit- and area-level SAE techniques implemented by the FAO, integrating survey data with additional auxiliary information retrieved from multiple potential data sources. In all the cases considered, the use of SAE techniques allows to increase estimates precision and produce predictions in out-of-sample domains.
Secondly, many large-scale surveys may not collect the variable needed for the estimation of the relevant SDG Indicator, but only auxiliary variables of general use. In this context - and in presence of a smaller specialized survey collecting the variable of interest along with some of the auxiliary information included in the larger survey - an indirect estimation approach, based on the so-called projection estimator, can be used to project synthetic values of the variable of interest in the larger sample and, hence, estimate the related SDG Indicator. Also in this case, the theoretical discussion of the advantages of the proposed statistical method is complemented by an empirical application.