Abstract
This paper concerns computational models in environmental economics andpolicy, particularly so-called integrated assessment models. For themost part, such models are simply extensions of standard neoclassicalgrowth models, extended by including the environment and pollutiongeneration. We review the structure of integrated assessment models,distinguishing between finite horizon and infinite horizon models, bothdeterministic and stochastic. We present a new solution algorithm forinfinite horizon integrated assessment models, relying on a neural netapproximation of the value function within an iterative version of theBellman equation.