Regularity of free boundary arising from optimal exercise of perpetual executive stock options (Q2515793)

The paper proves the regularity of a degenerate Stefan type free boundary problem that results from a variational inequality modeling the value function of a perpetual executive stock (American call) option. In particular the authors consider the value function \[ V(x,s,A) = \lim_{T\to\infty} \sup_{\{ A(t) \} \in \mathbb A } \mathbb E\left[ U\left( x - \int_0^T e^{-r t} [S(t) - K]^+ dA(t)\right) \Big| S(0) = s\right], \] where \(x,s,K,A\) are the initial cash, the stock price, the strike price, and the ESO holdings, \(r\geq 0\) is a constant discount rate and \(U\) is a given utility (concave increasing) function, and \(\mathbb A \) is the set of all the admissible holding strategies. By assuming a Black-Scholes model \[ dS(t) = \alpha S(t) dt + \sigma S(t) dW(t) \] for the stock, and an exponential utility function \(U(s) = -e^{-\gamma z }\), the authors show that the value function \(V\) can be written as \(V = U(x + \gamma^{-1} \varphi)\) where \(\varphi(z,a)\) is the viscosity solution of the variational inequality: \[ \min\left\{ R a \varphi_a - \varphi_{zz} - \nu \varphi_z + \varphi_z^2, \varphi_a - g^+ \right\} = 0\;\text{ in } \;\mathbb R \times (0,\infty),\;\varphi(\cdot, 0) = 0. \] Here subscripts represent partial derivatives, \[ z = \log \frac{s}{K},\;a = \gamma K A, \;R = \frac{2r}{\sigma^2}>0,\;\nu = \frac{2 \alpha}{\sigma^2} -1,\;g(z) = e^z -1. \] By defining \(\psi = \varphi_a\) and \(w = \varphi_{aa}\) and making the assumption that \(R > \max\left\{ \nu + 1, 0\right\}\), the authors derive the following Stefan type free boundary problem \[ \begin{cases} (R a w + 2 \varphi_z \psi_z)_a = w_{zz} + \nu w_z - Rw & \forall a \geq 0, z < s(a), \\ w = 0 & \forall a \geq 0, z < s(a), \\ w_z = [g''+(\nu-2\varphi_z)g'-Rg] s' & \forall a \geq 0, z < s(a)^-, \\ \psi_a = w, \varphi_a = \psi & \forall a \geq 0, z \in \mathbb R, \\ w(\cdot,0) = w_0, \psi(\cdot,0) = \psi_0, \varphi(\cdot,0) \;& \forall a \geq 0, z \in \mathbb R, \end{cases} \] where \(z = s(a)\) is the free boundary. The authors establish the regularity of the free boundary by solving directly the free boundary problem.