Using integrated data sources for small area estimation within italian social surveys to monitor SDG indicators
Conference
64th ISI World Statistics Congress - Ottawa, Canada
Format: IPS Abstract
Session: IPS 375 - Small Area Estimation for Sustainable Development
Monday 17 July 4 p.m. - 5:25 p.m. (Canada/Eastern)
Abstract
Over the last years, ISTAT has begun to develop studies for the estimation of the targets related to the SDGs
for finer domains than those consider by the 2030 Agenda using Small Area Estimation methods. The most
recent studies focused on indicators related to health, gender equality, and poverty. In this work the results
related to the study of the variable “At risk of poverty or social exclusion”, AROPE, are presented.
The AROPE rate is the share of the total population which is at risk of poverty or social exclusion. In Italy, it is
the main indicator used to monitor SDG 1 as well as the headline indicator used to evaluate whether the level
of social targets required in European countries by 2030 is met. It includes the people who are either At Risk
of Poverty (ARP), or SEVerely materially and socially DEPrived (SEVDEP) or living in a household with a very
Low Work Intensity (LWI). People are included only once even if they are in more than one of the situations
mentioned above. This indicator is measured by means of data coming from the European Survey on Income
and Living Conditions (EUSILC) and is released for the five macroregions, for ten age classes and provides
reliable estimates for Administrative Regions (NUTS2). The request of estimates at province and metropolitan
area level (NUTS3) is very high from stakeholders and, in particular, from those who promote and carry out
anti-poverty and social exclusion interventions. In this talk we describe the process of production of SAE
estimates for ARP and AROPE to meet this information gap using detailed administrative information such as
tax records. The talk illustrates the implementation of well-established SAE methods that use both area-level
and unit-level models and proposes a set of tools to help selecting the best method, with a focus on model
selection, model diagnostics and benchmarking.