Dynamic programming is the mathematical foundation for economic analysis of dynamic problems. This masterclass describes the latest developments in computational methods for solving large, complex dynamic programming problems. This includes value function iteration methods for life-cycle models and simultaneous equation methods for infinite-horizon problems that use new approaches to value function approximation to achieve both numerical stability and high accuracy. For example, we will solve finite-horizon life-cycle models including labor supply, consumption decisions, and dynamic asset allocation of six stocks, one bond, and proportional transaction costs for stock transactions.
We illustrate these methods by computing responses to alternative tax policies and showing how to incorporate these methods in structural estimation methods, such as maximum likelihood and method of moments estimators. We will discuss various implementations of these methods on distributed and parallel computing environments, often achieving 95% efficiency when using hundreds of processors.