Howard Marks Presidential Professor of Economics at @Penn and Senior Fellow at @AEI. Demographics, AI & macro.
All opinions are my own.
Apr 27, 2025 • 5 tweets • 4 min read
As the world debates new economic architectures, a reminder that some, like Michael Pettis (@michaelxpettis), have been spectacularly wrong before about international macroeconomics:
“Spain will almost certainly be forced to leave the euro and default on its debt, or else suffer many years of crippling unemployment, rising debt, and economic stagnation.”
As of tonight:
— Spain is still in the euro.
— Total employment is up 18.2% (2024.Q4), a historical high, vs. when article was written (2011.Q1).
— Our private sector is a net lender to the world: elconfidencial.com/economia/2025-…
something that is probably a first in our economic history.
— Real GDP has held up nicely.
Hard to be so wrong about so many things!
More fun posts to read from the perspective of time:
Deep learning (DL) is proving remarkably effective at solving dynamic models in economics:
📄
Curiously, DL often yields stable solutions without explicitly imposing boundary conditions (e.g., transversality).
Why does this work? A short thread 🧵👇
1/7nber.org/papers/w33117
In a recent paper 📄 , we argue this phenomenon comes from the inductive bias of first-order optimizers.
Deep neural networks (DNNs) are hugely overparameterized—many weight sets can perfectly fit the data (e.g., grid points in state space).
🧵 below and ungated version at sas.upenn.edu/~jesusfv/Defen…
1/10
Why has U.S. productivity growth slowed down despite record R&D spending and more researchers than ever?
We explore a surprising answer: incumbents are hiring researchers not to innovate—but to block competition.
Aug 24, 2024 • 6 tweets • 1 min read
@ModeledBehavior @pseudoerasmus No, the measure we use in our paper does NOT exaggerate Japan’s economic strength:
1) Most importantly, we are comparing growth rates, NOT income levels. Thus, what matters is whether labor participation rates of the elderly in Japan have changed since 1991.
1/n
@ModeledBehavior @pseudoerasmus 2) The participation rate of those over 65 in Japan in 1991 was 25.3%, and it is still 25.3% in 2019. No change! 3) So, Japan’s income level was higher in 2019 because its elderly work more, but it was already higher in 1991 because of the same reason. 2/n
Aug 9, 2024 • 5 tweets • 1 min read
@Noahpinion @King_ofSweden @nberpubs @OurWorldInData Let me re-write the responses in a more logical way:
1) One does not want to use PPP-adjustment values to compare growth rates across countries (as in the left panel). Even the PWT webpage discourages such use. 2) Regrettably, PWT PPP adjustments are quite unsatisfactory. 1/n
@Noahpinion @King_ofSweden @nberpubs @OurWorldInData 3)In the case of Spain, the PPP values contradict every other observation we have about the Spanish economy.
4)Hours data for Spain are particularly bad because of the large flow of immigration since the late 1990s.
2/n
