ทวีตที่ปักหมุด
Joe Fish
2.1K posts
Joe Fish
@sadBusdriver
@sadbusdriver.bsky.social
San Francisco, CA เข้าร่วม Temmuz 2012
428 กำลังติดตาม269 ผู้ติดตาม
@the_good_matty @alz_zyd_ I mean, the hypothetical mba student is an obvious caricature, but I think that it's only ex-ante dumb because of theory, which is my point
English
@sadBusdriver @alz_zyd_ sounds like you got a dumb student and you should feel bad for putting this guy in a position of power
English
Given how easy it is to estimate upward sloping demand curves that fit the data really well, I'm kind of surprised at the "AI will / can solve causal inference" takes.
E.g., a regression of changes in home values on changes in supply gets an R2 of 0.88 and the wrong sign!
English
@zyudhishthu @_brianpotter @Austan_Goolsbee I think this is too pessimistic! even having Canada's construction growth would be what a 40% increase compared to what happened in the US?
if their productivity measure is something like homes built per worker, that would show up as a huge decline in breakeven rents
English
A bit pessimistic from @_brianpotter: we know construction productivity growth has been awful in the US the past couple decades (cc @Austan_Goolsbee).
However, there's not really any rich, large country that has seen strong growth in construction productivity (except Belgium?)
English
@_vonarchimboldi @alz_zyd_ it's a literature, for sure, but afaict, nobody in economics really does it.
if I take the past ten issues of QJE and AER, how many papers would there be that do formal sensitivity analysis? ten? five? one? in contrast, it's way bigger in epidemiology
English
@sadBusdriver @alz_zyd_ They exist. See proceedings.mlr.press/v97/cinelli19a…,
papers.ssrn.com/sol3/papers.cf… and arxiv.org/abs/2112.13398
English
Causal inference is a subfield of prediction, where you try to predict what happens when a treatment is applied. So all of econometrics really is just a narrow subfield of predictive machine learning
Arpit Gupta@arpitrage
Causality and prediction are not two distinct concepts Causal inference is fundamentally a prediction problem: you’re predicting the counterfactual Randomistas found a few clever ways to do causality without prediction. But if you solve prediction; you get causality for free
English
@arpitrage I think not quite getting to "solve prediction" is a dangerous place to be in!
A kitchen sink regression on changes in rent prices as a function of baseline demographics + changes in housing units gets you an R2 of 0.86 and an upward sloping demand curve!
English
Causality and prediction are not two distinct concepts
Causal inference is fundamentally a prediction problem: you’re predicting the counterfactual
Randomistas found a few clever ways to do causality without prediction. But if you solve prediction; you get causality for free
English
@alz_zyd_ I found all the micro econ 101/201 stuff really, really helpful at explaining the world... after I had gone through a PhD IO course (basic-y insights like cournot can be Bertrand w/ capacity constraints).
not sure if this reflects good or bad on econ 201 though lol
English
I agree this is an issue with undergrad econ education. Interestingly, good finance courses do a better job with this. An undergrad investments class is both very practically useful for finance job interviews, and teaches some deep principles about the theory of financial markets
Jostein Hauge@haugejostein
As an undergraduate student, I left the traditional economics track for one simple reason: I wasn’t learning much about how the economy actually works in the real world. Most of my classmates and I spent our time solving abstract mathematical problem sets that had little connection to the economic issues we were reading about in the news. When friends outside the field asked me for insights into even the simplest economic questions, I struggled to give good answers. It deeply frustrated me. I’ll never forget when one professor openly told us that teaching us real-world economics “wasn’t their job.” I was stunned. It felt as if I had unknowingly signed up for a watered-down degree in applied mathematics. That moment pushed me to look beyond traditional economics departments for further study and explore more interdisciplinary approaches to understanding the economy. I’m forever grateful that I did.
English
@Noahpinion and the college wage premium has been pretty flat, so, kind of paradoxically, now seems like a very good time to be going to college! (plus or minus your opinion on AI and the labor force)
minneapolisfed.org/article/2025/w…
English
@Noahpinion beginning in 2020, college tuition started falling relative to CPI (after 40 years of outpacing it)
this is either because of weaker demand, or college is finally getting cheaper right as people are adjusting their expectations!
English
The U.S. higher education sector is only at the beginning of a long and painful reckoning
Sahil Kapur@sahilkapur
Poll: In a dramatic shift, Americans no longer see four-year college degrees as worth the cost nbcnews.com/politics/polit…
English
@besttrousers @Earl_of_Madness @hecubian_devil (but also Faircloth is not binding in any meaningful way).
Repealing it would be good, but mostly symbolic
x.com/sadBusdriver/s…
Joe Fish@sadBusdriver
Unfortunately, faircloth isn't what's holding back public housing. Reality is that most PHAs are 1000s of units under their cap. Bigger problems are money, NIMBY regulations, and stuff like California's Article 34, which makes all low-income public housing subject to a vote.
English
@Earl_of_Madness @hecubian_devil Sure, I support the repeal of Faircloth.
Seattle, WA 🇺🇸 English
Many things wrong with this (for one, Faircloth was 100% a NIMBY proposition), but wouldn’t repealing Faircloth be deregulation?
It’s a regulation that restricts housing construction! I was told deregulation is neoliberal and bad, and repealing rules is always libertarian.
Jerque Cousteau@neo_antiquarian
We can’t build new public housing. It’s not legal. Because the ancestors of YIMBY’s past did a thing called the Faircloth Amendment. 30+ year old Baristas shilling for market rate housing is giving Caitlyn Jenner for Trump. Go off I guess, Queen
English
@nominalthoughts like you'd have to believe China has lower extreme poverty than the following countries: Norway, Denmark, Sweden, Austria, the Netherlands, Finland, Italy, ...
English
@nominalthoughts china is reporting a 0% extreme poverty rate lol
but almost certainly what's happening is the world bank survey isn't capturing in-kind benefits in the US correctly.
bsky.app/profile/sadbus…
English
No joke, I think this is the worst/most deceptive graph I have ever seen in my life
Jostein Hauge@haugejostein
The share of the population living in extreme poverty is now higher in the United States than in China.
English
even in a competitive market, if demand for NYC tourism went up a lot, supply would respond, but unless supply is perfectly elastic, prices should go up.
of course, algo pricing + collusion can happen on top of this, but supply increased so must be collusion isn't correct
English
not trying to dunk, but I think this is a thing people get tripped up on
unless you think making hotels got cheaper (seems unlikely, but I genuinely don't know), the new hotels are responding to a shift in the *demand* curve and so supply should correlate positively with/ prices
Lee Hepner@LeeHepner
NYC’s hotel inventory increased 73% between 2007-2021. Demand is (presumably) inelastic. A big reason we can’t have a real conversation about prices is that every major chain - Hilton, Marriott, Wyndham, Hyatt, Omni, Choice Hotels, etc - is using the same price-setting software.
English
@isaacinthesky with any of these though, I think the effects end up being pretty constrained by the fact that hotels and houses are already built
so the economic incentive to withhold supply isn't very strong. eg real page moves occupancy rates from ~97->~95% and increases prices by a couple %
English
@isaacinthesky it's a question of whether you think the software maximizes joint or individual profit.
if it's individual, there are incentives to undercut, if it's joint, the algo can facilitate collusion (one piece of evidence is real page banned users who rejected price suggestions)
English
Can someone explain the obsession with price setting software among the anti monopoly left to me? You'd think if anything it'd make the market more effecient and therefore decrease prices
Lee Hepner@LeeHepner
NYC’s hotel inventory increased 73% between 2007-2021. Demand is (presumably) inelastic. A big reason we can’t have a real conversation about prices is that every major chain - Hilton, Marriott, Wyndham, Hyatt, Omni, Choice Hotels, etc - is using the same price-setting software.
English