Final Report Summary - POEMH (Parsimony and operator methods for treatment of endogeneity and multiple sources of unobserved heterogeneity)
The ERC POEMH focused on the development of nonparametric and high-dimensional for inference in Econometric models with endogeneity and multiple sources of unobserved heterogeneity. The goal pursued has been to propose methods to relax assumptions made for convenience (monotonicity, rank invariance, parametric assumptions, strong instruments and their exogeneity, etc.) and propose methods which use realistic assumptions on data sets (e.g. limited exogenous source of variation), are computationally attractive, with well-grounded algorithms rather than heuristics, data-driven and permit quantifying uncertainty. The main difficulty of such a research agenda is that we are at the verge of identification so that we can obtain less precise conclusions. For this reason, we have proposed sensitivity analysis (e.g. in the level of sparsity) and nested sets of assumptions which are general enough to contain (arbitrary) assumptions usually maintained. The applications have been carried on demand analysis and health and some work still in progress concernes development and international trade.