Model averaging (MA) estimators in the linear instrumental variables regression framework are considered. The obtaining of weights for averaging across individual estimates by direct smoothing of selection criteria arising from the estimation stage is proposed. This is particularly relevant in...