To be clear, not saying that you can never identify human capital or signaling effects (@caroartc has a great recent example of doing so). I'm saying that you can't figure out *how important those two theories are in relative importance*, which really limits how useful they are!
-
Show this thread
-
Figuring out whether ed returns are mostly-HC or mostly-S has been tried a lot of times, and nobody agrees. Perhaps because they can't actually identify it! I provide a causal model (yes, a DAG) showing what the ID of relative importance even means, and give partial-ID bounds.
1 reply 0 retweets 10 likesShow this thread -
Then I try to see how narrow we can get these partial-ID bounds. Turns out, not very! The bounds are heavily based on identifying mediated causal effects of education, and cleanly assigning those effects to one of two theories.
1 reply 0 retweets 3 likesShow this thread -
Basically, both theories are so "squishy" that none of the empirical content of them we have so far - employer learning, sheepskin effects, aggregate productivity - actually have indeterminate interpretation. I show this by looking at the literature, see paper for details!
2 replies 0 retweets 7 likesShow this thread -
What about the quasiexperiments? Those are better at identifying effects with clean theoretical interpretation for *one* theory, but by design they end up cutting themselves off from being able to say much about that one theory *relative to the other, or the whole return.*
1 reply 0 retweets 6 likesShow this thread -
This implies econ of ed should maybe think about a different theory for organizing its wealth of empirical info about returns (suggestions in paper!), and that traditional policy implications of HC vs. S don't hold anyway! The HC/S - external/private analogy doesn't hold up!
1 reply 0 retweets 10 likesShow this thread -
There's a lot going on in the paper so I can't cover every point in a Twitter thread, so do check out the paper if you're interested in HC/S or identification. Ungated version at http://nickchk.com/Huntington-Klein_2020_Human_Capital_vs_Signaling.pdf … and an old version with way less math at http://nickchk.com/Old_WP_Human_Capital_Signaling.pdf …
1 reply 0 retweets 6 likesShow this thread -
I think it's a very *cool* paper in terms of being unusual. Why? (1) it's an applied paper with no new data, (2) it's both lit review and new argument, (3) theory paper with no formal economic theory, (4) impossibility-proof that never gets to 100% impossible, just close.
2 replies 0 retweets 9 likesShow this thread -
Everything is framed in terms of the partial identification bounds, but under the hood it's a very textual argument about economic theory and identification, which feels very old-school to me.
1 reply 0 retweets 5 likesShow this thread -
Many thanks to EE who took a weird paper and definitely helped make it better, and
@dlmillimet for suggesting EE (he has now invited DAGs in his house, like a vampire). Tagging other people who might like this for its DAGs or for its content:@PHuenermund@causalinf@eigenrobot3 replies 0 retweets 12 likesShow this thread
buddy this owns and I can't wait to read it! Congratulations, I hope this gets recognized :)
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.