A geometric programming approach to dynamic economic models

Geometric programming (GP) has several attractive features: it is tractable in large-scale problems, requires no initial guess or tuning of solver parameters, guarantees the convergence to a global optimum and can deal with kinks. In this note, I argue that GP is a potentially promising tool in economics. First, I show that a stylized finite-horizon growth model can be mapped into a GP format by using simple transformations. Second, I show that GP methods produce accurate and reliable solutions including the case of occasionally binding constraints which cannot be easily treated with conventional solvers. Examples of MATLAB codes are provided.