Chapter 1 firstly documents that in the U.S., the stock market participation rate over the life-cycle decreases as people age. This fact, however, can not be captured by standard model where smooth expected utility function predicts the decision maker stay in the stock market, given positive equity premium and independence between older people's non-asset income and stock return. This paper successfully replicates this fact by introducing Knightian uncertainty into a dynamic model with choices on medical care and consumption, as well as saving and portfolio. We assume the agents in the model have ambiguity towards the
correlations between risky stock return and risky medical price inflation.
In this environment, retirees quit the stock market under reasonable range of ambiguity towards the correlation. The key mechanism is that: The agents do not hold positive amount of stocks since they worry stocks co-vary positively with the non-asset income minus the health expenses.
Similarly, they do not short stocks as they also worry that stocks and their de facto non-asset income may co-move negatively.
Chapter 2 investigates the nature of riskiness in borrowing costs and its implication for households' portfolio choice over the life cycle. This paper firstly documents a negative correlation between borrowing cost and stock return as well as a wedge between saving and borrowing rates. This fact is then incorporated into life cycle portfolio choice model as key ingredient to study households' risky asset, safe asset and debt holdings. The model can deliver that the households' mean assets holdings are non zero in all three categories: loans, safe bonds and risky assets. And it shows as relative risk aversion increases, the asset holdings among all categories become more realistic. This improves the life cycle models with tight borrowing cost, where loan holdings are zero, which is inconsistent with data; and it improves from the model with loose borrowing constraint and risk free borrowing cost as low as saving rate, where agents invest too much in stocks. Yet the model can't deliver realistic asset holding with low constant relative risk aversion.