Multivariate Small Area Estimation of Multidimensional Latent Economic Wellbeing Indicators

• Activity Number: 410 - Social Issues, Trends, Inequality, and Employment
• Type: Contributed
• Date/Time: Tuesday, August 1, 2017 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Angelo Moretti* and Natalie Shlomo and Joseph Sakshaug
• Companies: University of Manchester and University of Manchester and The University of Manchester
• Keywords: EU SILC ; Factor analysis models ; Latent variables ; Model-based inference ; Multivariate EBLUP ; Multivariate multilevel models

Abstract

In the study of multidimensional wellbeing phenomena, policy makers need reliable information about the geographical distribution of social indicators. Factor analysis (FA) models are models used in data dimensionality reduction problems where latent variables are estimated to reflect the concept of wellbeing. In the wellbeing measurement, we may expect that factors are naturally correlated (Thorbecke, 2008), hence multivariate models would be appropriate for the prediction of wellbeing in small areas. Datta et al., (1999) shows that the use of multivariate mixed effect models in small area estimation (SAE) might yield improvements in terms of the reduction in mean squared error compared to the univariate approach. We employ FA models and multivariate empirical best linear unbiased predictor (EBLUP) in order to predict a vector of means of factor scores representing wellbeing for small areas. We compare this approach to the standard approach whereby we use SAE (univariate and multivariate) to estimate a dashboard of EBLUPs on original responses and then averaged. A simulation study and an application using the EU Survey of Income and Living Conditions data demonstrate our approach.