POLS 208 Lecture Notes - Lecture 17: Anuja Iyer, Instrumental Variable
Document Summary
Continuity assumption: nothing changes at the cut point other than the treatment status, must be met for valid rd. Incumbents are not directly comparable to non-incumbents: partisan district, better politician, effective campaigner, regression discontinuity, compare incumbents who barely won the last election to candidates who barley lost the last election. Instrumental variables: goal: estimate causal effect of some treatment, by overcoming endogeneity problems, examples, corn prices (rainfall, acemoglu, johnson & robinson, estimate using 2sls, aka two-stage least squares, basically two regressions. Good instrument: valid (predict the iv, relevant, covers a lot of the variation, meets the exclusion restriction, only predicts the dv through the iv. Example: corn prices: do shifts in quantity cause shift in price, hard to determine because of endogeneity, supply and demand. Instrument for quantity: rainfall: direct effect on quantity of corn, valid instrument, affects quantity but not price, exclusion restriction. Acemoglu, johnson, & robinson (2001: dv: economic development, log gdp per capita in 1995.