Best of #econtwitter - Week of September 18, 2022 [3/3]
Welcome readers old and new to this week’s edition of Best of Econtwitter. Please submit suggestions — very much including your own work! — over email or on Twitter @just_economics.
This is part three of three.
Paper summaries
We worked with @LinkedIn to analyze a set of experiments on their People You May Know (PYMK) algorithm, which recommends new ties to @LinkedIn members. Members were randomly assigned to different variants of the algorithm that gave them more weak tie or strong tie recommendations
We then waited, for 5 years, as experimental subjects professional networks and job mobility outcomes evolved... To estimate whether making more weak tie connections (or strong tie connections) increased job mobility over time.
^another thread from coauthor: “moderately weak ties produced the greatest prospects of them finding new positions”
👉 Individuals who excelled in the International Mathematical Olympiads born in poorer countries are systematically less likely to engage in knowledge production.
"Invisible Geniuses: Could the Knowledge Frontier Advance Faster" by @RuchirAgarwal_& Gaule (2020, AER:I).
A new research study from @mattkahn1966 @RDMetcalfe @S_Olascoaga and myself found that @Redfin users who viewed homes with severe and/or extreme flood risk prior to a Redfin experiment proceeded to bid on homes with 54% less risk after gaining access to risk data. 2/11
"Superhuman science"
This paper explores the potential impact of AI on scientific fields where innovation is about searching big spaces of possibility (e.g. new drugs, materials, architectures to build better AI systems...).
brookings.edu/wp-content/upl…
#econai ht @mattsclancy
Not quite the results we were expecting (or hoping for!) but they are what they are...
Work lead by @iliasberberi 👏 (see 🧵⬇️)
OA version: osf.io/preprints/meta…
Paper: nature.com/articles/s4155…
1/ Do you remember the Spotify boycott in January and headlines like this in the media?
Did you wonder what happened to the actual Spotify revenues and subscriber numbers?👇
cc @tuchmanna @zhu_xinrong
#EconTwitter #MarketingTwitter
^nothing, nothing happened
Excited to share a new working paper #EconTwitter: what are the short- and long-term effects of cash for work in Tunisia? Joint w/Eric Mvukiyehe (formerly at WB DIME, now Duke)
It is very hard for startups to hire - even Google had to hire their landlord (by luck, the great Susan Wojcicki!). Hypothesis: it's hard b/c no one knows which startup is actually great. @AmirSariri , Mitch Hoffman & I looked into this in a new draft:
More: lemons and non-lemon real estate; childcare program eval; time-varying disaster risk; impact of drought on employment
Interesting discussions
Idea: If you publish a paper in a journal, the journal should let you add public comments about that article in a very low-cost way, so future readers can contextualize that work. It would extend the useful life of old papers.
We have spent decades trying our best to pretend that research excellence is somehow tied to the capacity to teach undergraduates. It has never made even a lick of sense; indeed, the opposite is more likely to be true.
Bret Devereaux @BretDevereaux
If you work for a major state university in the social sciences or humanities you are being paid to teach. And yet hiring and promotion decisions are almost completely indifferent to teaching.
And of course *everyone* knows this.
@mattyglesias @drfarls It's honestly insane, and at the extreme ends means dragging genius researchers away from vital work so they can completely fail at teaching, while driving fantastic educators to nervous breakdowns by gatekeeping their careers behind "publish or perish" mandates.
^John Cochrane says this would inevitably cause teaching to be outdated, but isn’t the obvious solution to have frontier researchers *teaching frontier courses* i.e. PhD courses — which is typically the desire anyway — and have specialized lecturers teaching the intro/intermediate courses, for mass consumption?
In Dec 2020 I created an animation of the FWL theorem. (Link to original post at end of thread)
I think it would be more clear why the x1 coefficient sign flips going from univariate (y ~ x2) to multivariate (y ~ x1 + x2) with residualization if the points were different colors.
Twitter has come down heavy on the notion that you get better feedback at a seminar if you haven't written the paper yet. But, what about distributing a paper for comments? Do we do that anymore? I thought I would do it for the seminar paper that I just finished for 1/2
John Whitehead @johnwhitehead81
