Essays in asset allocation with recursive utility and regimes in asset return

Approximate Analytical Solutions for Consumption and Portfolio Decisions under Recursive Utility and Finite Horizon: In the first chapter, we study the asset allocation and consumption decisions of an investor with stochastic differential recursive utility and a given  finite investment horizon. We provide an approximate analytical solution for this problem under a stochastic investment opportunity set. The solution becomes exact when the elasticity of intertemporal substitution parameter is equal to one or under a constant opportunity set. We show that the elasticity of intertemporal substitution parameter impacts both consumption and portfolio strategies, indicating the importance of disentangling intertemporal substitution from risk aversion. The investor's horizon also plays a crucial role in the optimal policies.

Optimal Portfolio Strategies in the Presence of Regimes in Asset Returns: In the second chapter, we also analyze investor's optimal strategies but here in a regime switching economy with unobservable states and predictability of risky asset returns. We develop approximate analytical solutions to the unconstrained dynamic problem and propose straightforward methodologies to build buy-and-hold strategies as well as portfolios with no short-selling. The approximation is shown to be accurate in a simple two-regime setting. While the portfolio policy depends strongly on the current state of the economy, the consumption-to-wealth ratio is roughly state-independent. In an application with large and small stocks and bonds, we show that buy-and-hold investors pay a very high cost for not timing the market. The costs of not allowing short-selling can be high, especially when the current regime is less uncertain. Predictability changes considerably the optimal portfolios. Hedging demands are negligible with regimes but no predictability, but can be important with predictability. On the other hand, the consumption-to-wealth ratio is not very impacted by the predictor.