Best of #econtwitter - Week of March 26, 2023: interesting tweets
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.
Interesting discussions
Citations are often "currency" for academics. While not perfect, they provide a meaningful measure of academic impact, are easy to obtain, and heavily influence academic promotions.
But how do we measure the broader impact of scholarly work? Some thoughts…
1/
@MichaelEddy @ipogadog This is a great paper, but to give credit to a key source of the underlying data they use let me add a cite to @overtonio:
@StatisticsFTW Yes, though we model TANF in California and not those states (for now--we're working on it in NY and TX). That explains California's increase and spike at the low end.
CA policyengine.org/us/household?f…
TX policyengine.org/us/household?f…
FL policyengine.org/us/household?f…
NY policyengine.org/us/household?f…
^🤢
If this single parent of two in California earns $10k, they take home $38k. If they earn $50k, they take home $49k.
That's a 73% marginal tax rate over a $40,000 earnings range. And that's without childcare and housing subsidies!
When I teach preference relations, one thing I always emphasize is the difference between indifference and incomparability, that is, between not not caring and being unable to decide.
It’s a very important distinction.
^file under: completeness axiom is overrated
I really like John's adaptation and I try to set goals like "write an article that is good enough to be published in X journal" rather than "publish this article in journal X". Your paper's quality is something you control (for the most part) whereas
John Horton @johnjhorton
@StefanFSchubert I also try to put a lot of effort into trying to make results as clear and simple as I possibly can, which I think maximizes long-run impact but I suspect hurts in the short run because it makes what you're doing seem not very impressive
If I could give a single piece of advice to new PhD students (in poli sci at least): work as hard as you can to write a complete, submittable solo research article by the time you start your 3rd year.
Even if it is never published, the experience you get is insanely helpful
This article in the WSJ starts by citing the number of job openings in JOLTS, then talks about a bunch of sketchy job listing practices but fails to note that these wouldn't (or at least shouldn't) be counted as job openings in JOLTS:
AI
Wake up!
In Jan of this year, @bryan_caplan bet @MatthewJBar that no AI would reliably score an A on his economics midterm exams before 2029.
Three *months* later, GPT-4 scores an A.
Somehow missed this the first time I looked, but GPT-4 got *significantly worse* at microeconomics after it was trained to tell you what you want to hear.
Today's fun GPT+Econ exercise: asking it to develop example code for structural estimation from a description + latex equations. The goal was to get it to give me a function that could estimate an approximate mixed-logit model. Step 1: ask GPT-4 how best to structure my prompt
Don't use ChatGPT for academic research. It creates fake citations to papers that don't even exist.
Instead, use Consensus — an AI-powered search engine designed for academics.
Ask it a question and it'll give you a summary of top 5-10 (real) published papers.
^haven’t tried this one yet honestly but
Journals and political endorsements
In 2020, Nature endorsed Joe Biden in the US presidential election. A survey finds that viewing the endorsement did not change people’s views of the candidates, but caused some to lose confidence in Nature and in US scientists generally
This, from the Editor of Science, is so disappointing.
Not because I embrace a naive separation between values and facts.
No: because it's a profound waste of the precious institutions of Nature and Science for them to focus on the same dumb political axis as everyone else.
Holden Thorp, Science EIC @hholdenthorp
I don’t think the purpose of academia is to achieve justice. The purpose of academia is to get closer to the truth
Doctors should treat a patient independently of their moral failings (they should cure a murderer); Academics shouldn’t cast moral judgements on their findings
^could put this tweet in the newsletter every other week
I appreciate that @Nature's editorial board responded to my research graciously and promptly. The discussion of my paper offered is thoughtful, but I find the counter-arguments and the conclusion unconvincing. (1/n)🧵
nature @Nature
AI safety
What regulatory options make sense to reduce risk from AI? My tentatively preferred option is to allocate at least $100 billion a year or more for rewards and grants for AI safety innovations, assessed by a board of relevant CEOs and AI researchers (i.e. people with inside info).
The problems you might have thought would take decades or centuries for AIs to solve are already being solved:
Leopold Aschenbrenner @leopoldasch
AI safety becoming mainstream
(Good thread from Yale econ professor)
Jason Abaluck @Jabaluck
Many are talking about the existential risk of AI, where humanity is “wiped out”.
I think more likely risk is AI used by corporations to supercharge the exploitation human biases, leading to extreme polarization and isolation.
Think Wall-E & Tokyo Ghost, not terminator. 1/4
Right, and to some extent, this is a risk that is not specific to AI. Drugs, sugar, alcohol have already destroyed some societies... while other have developed ways to control the dangers of these products and get the best of them.
Same thing here, we need to sort out new norms
Alex Imas @alexoimas
Public goods
Finding regional / sub-national data in 🇪🇺 is often challenging. I am compiling a list of databases with regional data from statistical agencies across Europe and the corresponding R packages. Please add to this Google Sheet if you know any: docs.google.com/spreadsheets/d…
Fin
This recollection of Ed Witten's early career by a friend from his college days is really something. It shows that even geniuses can meander and flounder quite a bit before making a mark.
The moral: if you're meandering and floundering, don't worry -- your goal-directed, child prodigy classmates might have been famous physicists in the past, but now they will likely be supplanted by super-intelligent AIs, albeit with an uncertain timeline.
What’s funny about this is that you can, without needing the labels, immediately tell which line must represent the most popular choice for academics seeking engagement with the world.
^but:
Twitter can be a big time sink and toxic sometimes, but it can also be useful and enriching. For instance, this week I'm going to a conference to present on a paper I wrote inspired by a @johnjhorton tweet!
Alvin Christian @_AlvinChristian
Out of the 20 most interesting thinkers *under the age of 45*, how many are currently employed by universities?
^🤔
