The credibility revolution in soc sci research is important but some humility is in order - DID models misspecified for years, RD estimates often noisy/underpowered and published selectively, and exclusion restrictions in IV models often highly dubious. (Why I like experiments!)
Very true! Experiments are hardly immune to bad methods and selective reporting, especially at tiny sample sizes of past (ugh). But arguably still far more robust as a method - the ATE relies on very weak assumptions compared to, e.g., IV/RD/DID
https://twitter.com/paulgp/status/1443566149254459397?s=20
Yes! This *is* a huge step forward.
https://twitter.com/matt_blackwell/status/1443615868123066376?s=20
• • •
Missing some Tweet in this thread? You can try to
force a refresh
