Multinomial-based small area estimation of labour force indicators

This paper deals with small area estimation of labour force characteristics like totals of employed, unemployed and inactive people and unemployment rates. Small area estimators of these quantities are derived from a multinomial logit mixed model with independent random effects on the categories of the response vector. The mean squared errors are estimated both by explicit formulas and by bootstrap methods. Two simulation experiments designed to analyze the behaviour of the introduced estimators have been carried out. Finally, an application to real data from the Spanish Labour Force Survey of Galicia is given.