Causal Inference Methods: Event Studies, DiD, RDD & IV
Tap to flip · Swipe to navigate
76 cards
Question: What question does an event study ask in its simplest form?
Answer: What happens around the event (e.g., before vs after a policy, announcement, or change)?
Question: How does difference-in-differences (DiD) extend the event-study logic?
Answer: By adding a credible comparison group and asking what changed for treated units after the event relative to what would have changed anyway.
Question: What are the two related uses of DiD mentioned?
Answer: 1) Estimate one average effect of an event or policy (classic 2-by-2 DiD). 2) Study event timing when different groups are treated at different dates
Question: What is the fundamental problem of causal inference described using potential outcomes?
Answer: For each unit there are two potential outcomes Y_i(1) and Y_i(0), but we can never observe both for the same unit at the same time, so the individual
Question: Why are naive comparisons between treated and untreated groups potentially misleading?
Answer: Because of selection bias: treated and control groups may differ before treatment (confounding), so observed differences may reflect pre-existing diff
Question: Give the hospital example illustrating selection bias.
Answer: Comparing health outcomes of people who went to hospital versus those who did not may show worse outcomes for hospital patients, but this likely refle
Question: What core question does DiD ask to help build a counterfactual?
Answer: What happened to treated units after the event compared with a credible control group?
Question: Why might a randomized controlled trial (RCT) not be feasible for evaluating some policies?
Answer: Because of ethical constraints (can't randomly deny treatment), high cost, implementation issues like non-compliance and spillovers, and limited exter
Question: What quasi-experimental opportunity arises when many policies happen at a sharp point in time?
Answer: A before-policy versus after-policy comparison (Before → After) that can be used to construct a quasi-experimental analysis if a credible counterfactu
Question: What is the naive before-versus-after estimator for a treated group?
Answer: Effect_BA^ = Ȳ_After - Ȳ_Before (the change in the treated group's average outcome over time).