A time series multilevel model is developed to produce monthly unemployment figures at a provincial level and quarterly figures at a municipal level. The model is formulated in an hierarchical Bayesian framework and fitted using MCMC simulations. The model borrows strength over time and space as well as from auxiliary claimant counts series. Regression coefficients for claimant counts are allowed to vary between municipalities and the regression coefficients can be time-dependent. Another way of including cross-sectional correlations is obtained by modelling the spatial effects among random domain intercepts and among random slopes for the regression coefficients. To allow for the diversity of municipalities and possibly volatile time-dependence, non-normally distributed municipal random effects and trend innovations are investigated by using global-local shrinkage priors.