A Parameter Estimation Method to Minimize Instabilities in System Dynamic Models

This paper introduces a new method that facilitates the stability analysis of system dynamics models. The method is based on the concepts of asymptotic stability and Accumulated Deviations from Equilibrium (ADE) convergence. We prove several theorems that show that ADE convergence of a state variable will make its trajectory approach asymptotic stability. Achieving ADE convergence requires the solution of a policy optimization problem. We use an approach called Behavior Decomposition Weights (BDW) to reduce the search space associated with that optimization problem. We also demonstrate this method on two examples: a linear “inventory-workforce” model and a non-linear “mass business cycle model”. These examples illustrate the features of this method and the potential for the development of efficient tools to improve the quality of the optimization policies.