How to find true cause and effect in a world of bias
When data is messy and randomization is imperfect, how can researchers distinguish real impact from mere coincidence? This video explores Instrumental Variables, a powerful econometric tool used to uncover causal truths even when selection bias threatens to distort the results.
In studies of American school quality, it is often difficult to tell if high test scores stem from excellent teaching or simply a fortunate location in a wealthy neighborhood. This problem of selection bias is a central challenge in econometrics.
While school lotteries offer a way to randomize access, the results remain imperfect because families choose whether or not to accept the seat. Using the framework of Instrumental Variables (IV), MIT's Josh Angrist demonstrates how researchers can navigate this incomplete randomization. The presentation covers essential terminology—including the first stage, second stage, instrument, and reduced form—and examines the three critical assumptions required for success: the exclusion restriction, the independence assumption, and a substantial first stage.
Source: Introduction to Instrumental Variables (IV)
