A New Solution Method for Singular Stochastic Control Problems∗†

We consider a singular stochastic control problem, which is called the Monotone Follower Stochastic Control Problem and give sufficient conditions for the existence and uniqueness of a local-time type optimal control. To establish this result we use a methodology that has not been employed to solve singular control problems. We first confine ourselves to local time strategies. Then we apply a transformation to the value (i.e. the total reward accrued by reflecting the diffusion at a given boundary) corresponding to a particular local time strategy and show that it is linear in its continuation region. Now, the problem of finding the optimal boundary becomes a nonlinear optimization problem: The slope of the linear function and an obstacle function need to be simultaneously maximized. The necessary conditions of optimality come from first order derivative conditions. We show that under some weak assumptions these conditions become sufficient. We also show that the local time strategies are optimal in the class of all monotone increasing controls. As a byproduct of our analysis, we give sufficient conditions for the value function to be C on all its domain. We solve two dividend payment problems to show that our sufficient conditions are satisfied by the examples considered in the mainstream literature. We show that our assumptions are satisfied not only when capital of a company is modeled by a Brownian motion with drift but also when we change the modeling assumptions and use a square root process to model the capital. We also sketch how our approach can be generalized to Bounded Variation Follower Stochastic Control Problem and give the characterization of the optimal boundaries. We solve two inventory management problems. In the first problem the contents of the inventory fluctuate as a Brownian motion with drift and in the second one the contents fluctuate as an Ornstein-Uhlenbeck process.