Course Catalog

A recent econometric literature has clarified key conceptual differences between 'design-based' identification strategies, which leverage a specification of counterfactual exogenous shocks, and strategies that instead leverage restrictions on model unobservables (e.g. the popular 'parallel trends' restriction). This course will review this literature and its main practical implications for empirical work. Topics will include design-based identification with formula (e.g. shift-share) instruments and treatments, negative weights in regression analysis, tricks to characterizing instrument compliers, and how a design-based approach can relax identifying assumptions in structural estimation.

The course consists of three days of lectures and sessions for the students to present their own research.

Visit our website for more information (see course details).