William Chen

Economics PhD student at MIT | former research analyst on the FRBNY DSGE Team | macro, finance, macro-labor | he/him

[William Chen @chenwilliam77]

@oliverwkim @PatrickAdams21 and @SunYongKim1 have constructed HFI series for some euro area countries. I also think @_SiLviA_mA has a series for the UK.

[William Chen @chenwilliam77]

@vit_nwu For time series identification and the relationship with cross-sectional methods, see Chris Wolf’s syllabus and slides for Advanced Macro 14.461/462: https://t.co/2IMSqlEIoB

[Liam Bright @lastpositivist]

This paper by @fabianbgng is one of those ones where I think it could and should be field changing. Let me explain.

https://t.co/03xK1zg9dx

[William Chen @chenwilliam77]

@momin_rayhan @PatrickAdams21 Neural nets are like an improved version of ergodic set methods (eg Maliar, Maliar, and Judd (2011) JEDC). You’re still solving for the exact solution, but the way you update the interpolant is different (eg not by solving a system of nonlinear eqns as projection methods do)

[William Chen @chenwilliam77]

@momin_rayhan @PatrickAdams21 The point of the simulation step is to be an efficient “grid” constructor. In a statistical sense, you only need your interpolant solution to be accurate over the model’s ergodic set. It doesn’t matter to be accurate in zero probability states.

[William Chen @chenwilliam77]

@momin_rayhan @PatrickAdams21 Jaggedness would only result if for some reason the eqm conditions generate very small errors when the interpolation is jagged in a region despite the true solution being smooth. But this usually won’t be the case, and if it is, then the jaggedness doesn’t matter anyway

[William Chen @chenwilliam77]

@momin_rayhan @PatrickAdams21 That’s actually the wrong intuition. The closer the points are together, the smoother it will be because the interpolation is more flexible. Think about an Aiyagari model with a binding borrowing constraint. You want more points there due to high curvature. Same applies to NN

[William Chen @chenwilliam77]

@momin_rayhan Also Kekre, Lenel, Mainardi (2022) for a term structure model with 4 states and Di Tella and Hall (2022) with 3 state models using Smolyak grids (2/2)

[William Chen @chenwilliam77]

@momin_rayhan Continuous time: less out there but Schaab and Zhang (2022) released a toolbox for adaptive sparse grids and finite differences, which is used in Schaab’s JMP to globally solve a HANK model (1/2)

[Ivan Werning @IvanWerning]

Inflation, what is its most proximate cause? What is the right general framework to think about this messy phenomenon?

Very excited to put out this new paper exploring a generalized "conflict" perspective as the right answer.

Paper here: https://t.co/rfTCZ2GEt9

🧵a thread...

[William Chen @chenwilliam77]

@itsafronomics Plot and message are not novel, but the execution of the idea is entertaining.

[steven t. piantadosi @spiantado]

Yes, ChatGPT is amazing and impressive. No, @OpenAI has not come close to addressing the problem of bias. Filters appear to be bypassed with simple tricks, and superficially masked.

And what is lurking inside is egregious.

@Abebab @sama
tw racism, sexism.

[Jon Steinsson @JonSteinsson]

It is widely believed that good monetary policy must satisfy the “Taylor Principle” (i.e., that nominal interest rates should rise more than one-for-one with inflation). In important cases, I think this is a misconception. 1/

[Lukas Freund @_LukasFreund_]

Such a neat paper by @pontus_rendahl:

"Continuous vs. Discrete Time: Some Computational Insights"

🔗https://t.co/xh9K2Upfam

[William Chen @chenwilliam77]

@fabolange For top 5-10 places in 2021, all the offers I got had health insurance and at least $36,000, with several above $40,000.

[William Chen @chenwilliam77]

@Schoefer_B Caballero and Simsek’s forthcoming AER explains gradualism as optimal policy when markets and the Fed disagree: https://t.co/VVp7Roic7L