Institutions which publish macroeconomic forecasts usually do not rely on a single econometric model to mechanically generate their forecasts. The combination of judgements with information from different models complicates the problem of characterizing the predictive densities. This paper proposes a flexible (yet parametric) approach to estimate the joint and marginal densities of macroeconomic forecasting errors, combining judgements with sample and model information. We assume that the relevant variables have a multivariate normal skewed distribution, belonging to a class of distributions recently suggested by Ferreira and Steel (2007a, 2007b). Our method is less informal than the original procedure used by the Bank of England to generate its fan charts and it does not suffer from the practical limitations of other approaches available in literature.