Uncertainty about trends and variability in mortality and fertility can often be captured by stochastic models, and then used to make probabilistic projections of population size and structure. These projections can be used directly to illuminate the risks attendant on some policy choices - I will illustrate these for the question of whether and how much we should raise retirement age to help with Social security, in the US and elsewhere. I will also discuss problems with tying such uncertainty into dynamic economic models.