Impulse control of proportional reinsurance with constraints (Q638026)

The main object of the paper is the following controlled stochastic process \[ X_t = x + \int_0^t \mu u(s) ds + \int_0^t \sigma u(s) dW_s - \sum_{n=1}^\infty I_{\{\tau_n<t\}}\xi_n, \] where \(\mu>0\), \(\sigma>0\) are some parameters, and the control \(\pi\) consists of an adapted process \(\{u(t)\in[0,1],t\geq 0\}\) together with sequences of stopping times \(\{\tau_i,i\geq 1\}\) and non-negative random variables \(\{\xi_i,i\geq 1\}\). The authors consider the problem of maximization of the following performance function: \[ J(x,\pi) = E\left[\sum_{n=1}^\infty e^{-\lambda \tau_n} (-K+k\xi_n)I_{\{\tau_n\leq \tau\}}\right], \] where \(\tau\) is the first moment when the process \(X\) becomes non-positive, and \(\lambda>0\), \(K>0\) and \(k\in(0,1)\) are some parameters. Some natural constraints are imposed on the control. Such a problem can be regarded as the problem of finding optimal proportional reinsurance and dividend policy under both proportional and fixed transaction costs. The authors show that the optimal dividend policy is some band policy.