lukestein’s avatarlukestein’s Twitter Archive—№ 7,978

    1. I liked Jen’s whole thread about causal inference in sentencing length Among other things, it highlights a problem in MANY settings where some heterogeneous “policy” brings units from a_i to b_i = a_i + d_i (Here sentence length drives age at sentencing → age at release) 1/2 @jenniferdoleac/1256222315089727492
      Permalink On twitter.com Twitter logo lukestein’s avatar ♻️ 5 Retweets ❤️ 2 Favorites Mood -6 🙁
  1. …in reply to @lukestein
    We’re often interested in effects of variation in d. But if b matters for outcomes, we don’t have enough degrees of freedom to also allow heterogeneous effects across a! (Without e.g, functional form assumptions.) 2/3
    On twitter.com Twitter logo lukestein’s avatar ♻️ 3 Retweets ❤️ 2 Favorites Mood +6 🙂
    1. …in reply to @lukestein
      I’m sure I’ve forgotten or never knew a better way of describing this formally. But this comes up a lot in applied micro and corporate finance, where (as @Andrew___Baker notes) we sometimes see too much naive controlling for pre-shock levels of things like M/B ratios. 3/3
      Permalink On twitter.com Twitter logo lukestein’s avatar ♻️ 3 Retweets ❤️ 2 Favorites Mood +1 🙂