This intensive course, part of the 'math+econ+code' series, is focused on conceptual and computational toolbox for solving and estimating large matching models with transferable utility. It shows that when unobserved heterogeneity is Gumbel, the model collapses to the classic Choo–Siow setup, which is especially convenient: key objects have closed forms and estimation is straightforward. With more general heterogeneity however, that convenience disappears. Following Galichon & Salanié (2022), estimation boils down to solving optimal transport problems with nonstandard (generalized) entropic regularization. • We’ll see how simulation turns these problems into very large linear programs, and why the Dantzig–Wolfe decomposition is a natural workhorse for handling them at scale. Students will have the opportunity to write their own code (in Python), in the spirit of cooking lessons. There are no prerequisite other than the equivalent of a first-year graduate sequence in econ, applied mathematics or other quantitative disciplines. 

This course is very demanding from students, but the learning rewards are high.