The monthly unemployment rate is based on the data of the Dutch Labour Force Survey. In this paper a structural time series model is developed and applied to estimate the monthly unemployment rate for six domains. The estimation results under this model are compared with the generalized regression estimates and the estimates under some simpler models. With univariate structural time series models, information from other time periods is borrowed to improve the precision of the estimates. Further improvements are possible by borrowing information from other domains in a multivariate structural time series model. It turns out that the trends of the six domains are cointegrated. Only three common trends have to be estimated, which means that the information from other domains is used in an efficient way. Further improvements can be achieved by modelling outliers and by modelling the seasonal component in an efficient way. The standard error of the estimates is approximately halved by the time series approach, compared to the generalized regression estimator.