There has been a recent upsurge of interest in probabilistic population projections. Two methods have been suggested in the literature for forecasting the inputs into those projections: (1) a random lines (RL) approach, and (2) a simple autoregressive approach (AR(1)). The purpose of this paper is to study analytically the differences produced by the two methods. We do this in the context of a model of variability in population growth rates. Two cases are considered: One where there is no population age structure and one where there is one. In both, we find that the variance using the AR(1) approach is initially smaller than that of the RL approach, but that over time the variance using the RL approach grows more rapidly leading to an instant of time when the variances are equal.