In this paper we argue that once one departs from the simple classroom example, or `stripped down life-cycle model,' the empirical model for consumption growth can be made flexible enough to fit the main features of the data. More specifically, we show that allowing demographics to affect household preferences and relaxing the assumption of certainty equivalence can generate hump-shaped consumption profiles over age that are very similar to those observed in household-level data sources, without appealing to alternative explanations (such as liquidity constraints, myopia or mental accounting). The hump-shape is partly attributable to precautionary savings, and partly due to demographics; the tracking (whereby consumption jumps with income) is instead due to the permanent nature of the income shocks. We use US household-level data to estimate preference parameters and income profiles, and then simulate consumption profiles for different education groups. Our simulated profiles show that the key features observed in the data can be closely matched in simulation. We also show that neglecting uncertainty produces consumption profiles that are `too flat,' whereas neglecting demographics generates consumption profiles that peak `too late.'
