【AP】On robust mean variance portfolios(2)

Pre Link

On robust mean variance portfolios(1)

Robust MV model

Robust MV model as follows, which we refer to the problem as RMVP2:
$$\begin{gathered} \underset{x}{\max }\underset{r\in U_r}{\min }{r}^{T}x+(W_0-e^{T}x)R \\ s.t.x^T\Sigma x\le t^2 \end{gathered}$$
for an appropriately selected positive number $t$, the model can be reformulated into
$$\begin{gathered} \underset{x}{\max }{\hat{r}}^{T}x+(W_0-e^{T}x)R-\gamma \Vert {\Sigma }^{1/2}x{\Vert }_2 \\ s.t.x^T\Sigma x\le t^2 \end{gathered}$$
or equivalently based on the equation ${\hat{r}}^{T}x-e^{T}xR={\hat{\mu }}^{T}x$
$$\begin{gathered} \underset{x}{\max }{\hat{\mu }}^{T}x+W_{0}R-\gamma \Vert {\Sigma }^{1/2}x{\Vert }_2 \\ s.t.x^T\Sigma x\le t^2 \end{gathered}$$
Proposition 2: If $H\ge \gamma$ then RMVP2 admits the unique optimal solution
$$x^{\ast }=\frac{t}{H}{\Sigma }^{-1}\hat{\mu }$$
If $H<\gamma$ then it is optimal for an RMVP2 investor to keep all initial wealth in the riskless asset.
Proof: the Lagrange function using a non-negative multiplier $\lambda$:
$$L(x,\lambda )={\hat{\mu }}^{T}x+W_{0}R-\gamma \Vert {\Sigma }^{1/2}x{\Vert }_2+\lambda (t^2-x^T\Sigma x)$$
the optimal solution $x^{\ast }$ can be obtained according to the first-order condition and Slater condition
$$x^{\ast }=(\frac{\sigma }{2\sigma \lambda +\gamma }){\Sigma }^{-1}\hat{\mu }$$
The definition of $\sigma$ reveal that
$${\sigma }^2=\frac{{\sigma }^2}{(2\sigma \lambda +\gamma {)}^2}{H}^2$$
Assuming the constraint to be active and observing that $\sigma =t$, we have
$$\lambda =\frac{H-\gamma }{2t}$$
which is non-negative provided that $H\ge \gamma$. If $H<\gamma$ then the only feasible choice for $\lambda$ is zero along with a dual objective function value equal to $W_{0}R$ which is attained in the primal by a riskless portfolio, i.e. $x^{\ast }=0$.
The robust MV portfolio is identical to the MV portfolio obtained as a solution to the problem
$$\begin{gathered} \underset{x}{\max }{\hat{\mu }}^{T}x+W_{0}R \\ s.t.x^T\Sigma x\le t^2 \end{gathered}$$
That is, the investor maximizing robust expected return under a variance constraint makes a MV portfolio choice when his/her confidence in the estimation of the mean is high, i.e. $\gamma$ is smaller than $H$.(这种情况下对均值的置信程度很高).

Corollary 1:
Let $H\ge \gamma$.Then

1. choosing a maximum variance $t=\frac{T-W_{0}R}{H-\gamma }$, the RMVP2 investor holds an optimal portfolio identical to the RMVP1 investor with a target wealth equal to $T$.
2. choosing a minimum target wealth equal to $T=W_{0}R+t(H-\gamma )$ the RMVP1 investor holds an optimal portfolio identical to the RMVP2 investor with a variance cap equal to $t^2$.

Consider the following problem
$$\underset{x}{\max }\underset{r\in U_r}{\min }{r}^{T}x+(W_0-e^{T}x)R-\frac{\rho }{2}{x}^T\Sigma x$$
where $\rho$ is a positive scalar. Expanding the inner $min$ problem from the point of worst-case, we have
$$\underset{x}{\max }{\hat{r}}^{T}x+(W_0-e^{T}x)R-\gamma \Vert {\Sigma }^{1/2}x{\Vert }_2$$
which refer to as RMVP3.

Proposition 3: If $\gamma \rho <H$ then RMVP3 admits the unique optimal solution
$$x^{\ast }=(\frac{H-\gamma \rho }{\rho H}){\Sigma }^{-1}\hat{\mu }$$
If $\gamma \rho >H$ then it is optimal for RMVP3 investor to keep all initial wealth in the riskless asset.
Proof: The function is strictly concave.(RMVP3问题是非凸优化), the first-order necessary conditions yield the candidate solution:
$$x=(\frac{\sigma }{\gamma +\sigma \rho }){\Sigma }^{-1}\hat{\mu }$$
Developing the right-hand side, we obtain a quadratic equation
$${\rho }^2{\sigma }^2+2\gamma \rho \sigma +{\gamma }^2-H^2=0$$
with the positive root
$${\sigma }_{+}=\frac{H-\gamma \rho }{{\rho }^2}$$
provided that $\gamma \rho <H$.