Small area estimation and prediction of unemployment duration mean in Iran and its effect on the provinces using multilevel models

The need for small area estimates via survey data has long been recognized. The importance of using such estimates is that the sample size in some of small areas may be small or even zero so that direct estimates cannot be made with acceptable precision in each small area. If data within small areas have hierarchical structure، multilevel models should be used to make estimates with acceptable precision. In this paper، the labor force survey data of 8 different Iranian provinces، collected by statistical center of Iran in 1388 (Iranian solar year)، have been analyzed. In this study، response variable that is unemployment duration of the persons in the survey، is modeled according to auxiliary variables such as age، sex، connection with householder، education situation، marital status، educational level and previous labor experience. Since data structure is hierarchical، we used three-level models for modeling the response variable. In this modeling، the education situation and marital status variables are not significant. The fixed effects and variance components are also estimated by restricted maximum likelihood method and the random effects are predicted by the BLUP method and then the main results are reported and interpreted. Finally using this model، the small area estimations of unemployment duration for 8 different provinces are predicted.