1: \begin{abstract}

2: 	This paper investigates

3: 	dynamic and static fund separations and their stability for long-term optimal investments under three model classes. An investor maximizes the expected utility with constant relative risk aversion under an incomplete market consisting of a safe asset, several risky assets, and a single state variable. The state variables in two of the model classes follow a 3/2 process and an inverse Bessel process, respectively. The other market model has the partially observed state variable modeled as an Ornstein-Uhlenbeck state process.

4: 	We show that the dynamic optimal portfolio of this utility maximization consists of $m+3$ portfolios: the safe asset,  the myopic portfolio, the \textit{m} time-independent portfolios, and the intertemporal portfolio.

5: 	Over time, the intertemporal portfolio eventually vanishes, leading the dynamic portfolio to converge to $m+2$ portfolios, referred to as the static portfolio.

6: 	We also prove that the convergence is stable under model parameter perturbations.

7: 	In addition, sensitivities of the intertemporal portfolio with respect to small parameters perturbations also vanish in the long run.

8: 	The convergence rate for the intertemporal portfolio and its sensitivities are computed explicitly for the presented models.

9: \end{abstract}

10: