J. ltal.Statist. Soc. (1998) I. pp. 93-109 SMALL AREA ESTIMATION AT PROVINCIAL LEVEL IN THE ITALIAN LABOUR FORCE SURVEY Piero Demetrio Falorsi, Stefano Falorsi* National Statistical Institute, Italy Aldo Russo University Roma Tre, Italy Summary With regard to the Italian Labour Force Survey (LFS), this work presents the results of research project aiming to find estimation methods which improve the reliability of the estimates at the level of provinces. The 95 provinces are planned domains, within geo- graphical regions, for which separate samples have been planned, designed and select We consider the following estimators: post-stratified ratio, Fay-Herriot and three time series estimators which generalize the Fay-Herriot estimator. Furthermore, we develop two empirical analyses: the first is a comparative analysis by means of a Monte Carlo simulation from a real time series obtained from bi-annual data of LFS; in the second analysis we develop a comparative study of the estimators using real bi-annual survey data and the estimates of MSE of estimators. Keywords: Small Area Estimation; Time Series Approach; Kalman Filter; Standard Er- ror. 1. Introduction In Italy, as in many other countries, there is a growing need for current and reli ble data on small areas. This information need concerns most sample surveys realized by the Italian National Institute of Statistics (ISTAT), especially the La- bour Force Survey (LFS), which has been studied to warrant accuracy only for national and regional estimates. In the.past, ISTAT's solution to this problem was to broaden the sample with- out changing the sample estimation method. In the last few years, however, in order to find a solution to the non desirable aspects of oversized samples, re- search has been launched to identify estimation methods that may improve the * Address for correspondence: Servizio Studi Metodologici, ISTAT, Via De Pretis 74BN, 00186 Roma, E-mail: stfalors@istat.it 93