Learning, ambiguity and life-cycle portfolio allocation

In the present paper I develop a life-cycle portfolio choice model where agents perceive stock returns to be ambiguous and are ambiguity averse. As in Epstein and Schneider (2005) part of the ambiguity vanishes over time as a consequence of learning over observed returns. The model shows that ambiguity alone can rationalize moderate stock market participation rates and conditional shares with reasonable participation costs but has strongly counterfactual implications for conditional allocations to stocks by age and wealth. When learning is allowed, conditional shares over the life-cycle are instead aligned with the empirical evidence and patterns of stock holdings over the wealth distribution get closer to the data.