POLS 208 Lecture Notes - Lecture 15: Regression Discontinuity Design, Causal Inference, Covariate
Document Summary
Rd design is for differences with very narrow margins. Reasonable to argue that those right below the cutoff are very similar to those right above. You have to assume they"re similar **by chance. Once you reach 11, there"s a jump in the probability of receiving an award. Observed outcomes as a function of test scores. There"s no difference between points in the bottom two lines. **the jump is the causal effect in receiving an award to getting a scholarship for. Potential and observed outcomes in rd designs college. Assuming we can use observed control group as a proxy of unobserved control group. Observed treatment group as a proxy of unobserved treatment group. The basic intuition: the cut off point. The assignment ot treatment is determined either completely or partly by the value of a covariate, xi, being on either side of a fixed threshold.