We propose a density combination approach featuring combination weights that depend on the past forecast performance of the individual models entering the combination through a utility-based objective function. We apply this model combination scheme to forecast stock returns, both at the...