Douglas Miller (Cornell University)

Abstract

In this project we show how the Local Projections method of Jorda (2005) can be paired with quasi-experimental RHS specifications to estimate dynamic causal treatment effects.  We compare this approach with traditional alternatives used by microeconomists (such as distributed lag models and event study models), and highlight some advantages and robustness of the local projections approach.  We identify a short-panel (1/T) bias problem for the Local Projections approach, and present a novel bias-correction method that results in unbiased estimates.  Our applications include dynamic impacts of local business cycles on mortality, and impacts of the minimum wage on employment. (And maybe other applications still to be determined?)