Comparing numerical methods for solving the competitive storage model

This paper compares numerical methods for solving the competitive storage model. Because storage implies a nonnegativity constraint on stocks, the solution methods must be considered carefully. The model is solved using value function iteration and several projection approaches, including parameterised expectations and decision rules approximation. In considering a storage model with convenience yield, in which the inequality constraint is smoothed, perturbation methods are also applied. Parameterised expectations approximation proves to be the most accurate method, whereas perturbation techniques are shown inadequate for solving this highly nonlinear model. The endogenous grid method allows rapid solution if supply is assumed to be inelastic.