CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Multivariate small area estimation of educational poverty with latent variable models Authors:  Gaia Bertarelli - Sant'Anna school of Advanced Studies (Italy) [presenting]
Maria Giovanna Ranalli - University of Perugia (Italy)
Monica Pratesi - University of Pisa (Italy)
Abstract: Educational Poverty (EP) for young adults can be read as a deprivation of opportunities and rights related to culture, participation, environment, and social relations. It means being excluded from acquiring the skills needed to live in a world characterized by a knowledge-based economy and innovation. It is a latent trait, only indirectly measurable through a collection of observable variables and indicators purposively selected as micro-aspects, contributing to the latent macro-dimension. EP is measured in Italy by the Educational Poverty Index. A problem with this index is that it is based on direct estimates, which are reliable only at a regional level, while to intervene in the phenomenon, it is important to obtain information at a finer geographical level. This problem has been overcome by considering estimates based on a Fay and Herriot model in the aggregation process. However, none of the proposed indicators considers the true latent nature of the phenomenon. We aim to go beyond this limit and develop a new multidimensional indicator at a small area level, which is based on a unit-level latent variable model in order to capture the underlying hidden dimensions of EP from a set of binary manifest indicators. The proposed model is applied to data from the aspects of everyday life survey in Italy focusing on Provinces and on suburbs in Italian Regions.