In this paper we propose a parametric block wild bootstrap approach to compute density forecasts for various types of mixed-data sampling (MIDAS) regressions. First, Monte Carlo simulations show that predictive densities for the various MIDAS models derived from the block wild bootstrap approach...