A Toolkit for Optimizing Functions in Economics

Optimization algorithms must be among the most common numerical methods used by economists. Yet, there is surprisingly little guidance on choosing the appropriate one. This problem is most notable with regard to conventional versus global optimizers. Typically, a global optimizer is used when a conventional one fails after substantial ``fiddling'' with a conventional optimizer. This paper introduces three different, easy-to-use, tools (cross-sections, radius plots, and a measure of the non-quadratic behavior of a function) that are designed to indicate when a global optimizer is needed. With their use, researchers should spend less time fiddling and more time generating results.