CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1589
Title: Time series small area estimation for unemployment rates using latent Markov models Authors:  Maria Giovanna Ranalli - University of Perugia (Italy) [presenting]
Gaia Bertarelli - University of Milan - Bicocca (Italy)
Francesco Bartolucci - University of Perugia (Italy)
Abstract: In Italy, the Labour Force Survey (LFS) is conducted quarterly by the National Statistical Institute (ISTAT) to produce estimates of the labour force status of the population, at national, regional (NUTS2) and province (LAU1) levels. In addition, ISTAT also disseminates yearly LFS estimates of employed and unemployed counts and rates at a finer level given by Local Labour Market Areas (LLMAs). LLMAs are aggregations of municipalities and are defined at every census in terms of daily working commuting flows. In contrast with the NUTS3 and LAU1 levels, LLMAs are unplanned domains. The continuous nature of LFS allows us to borrow strength not only from other areas but also over time. We develop a new area-level SAE method using latent Markov (LM) models in a Bayesian setting. LM models allow for the analysis of longitudinal data when the response variable(s) measure common characteristics of interest that are not directly observable. In these models the characteristics of interest, and their evolution in time, are represented by a latent process that follows a Markov chain, so that statistical units are allowed to move between the latent states during the observation. Estimation is conducted using a Gibbs sampler with data augmentation and the proposed model is applied to estimate annual employment and unemployment rates for the Italian LLMAs using data from 2004 to 2013.