In this paper we present a framework for incorporating uncertainties into economic activity forecasts for Croatia. Using the vector error correction model (VECM) proposed by Rukelj (2010) as the benchmark model, we forecast densities of the variable of interest using stochastic simulations for incorporating future and parameter uncertainty. We exploit the use of parametric and non-parametric approaches in generating random shocks as in Garrat et al. (2003). Finally we evaluate the results by the Kolmogorov-Smirnov and Anderson-Darling test of probability integral transforms. The main findings are: (1) the parametric and the non-parametric approach yield similar results; (2) the incorporation of parameter uncertainty results in much wider probability forecast; and (3) evaluation of density forecasts indicates better performance when only future uncertainties are considered and parameter uncertainties are excluded.