Nathan Williams

Pleased to have passed my dissertation proposal! Thankful to have a community of scholars supporting me (esp. my advisor, Nels) and excited for the next steps of the PhD journey.

@FelixTintelnot Thank you for writing it! It’s been really helpful for some overfitting issues in my work on tract-level diffusion and growth. Looking forward to citing it when I go on the market next year.

@NathanEcon Great to hear, thanks!

One of my favorite papers that has shaped the way I think and the way I am writing my JMP.

@JohnRuf6 I bet it must have been really cool (and a great training exercise) to work on this paper. My tentative title for my JMP is “Granular Spatial Growth”, so you can tell how much I enjoy this paper

@NathanEcon hehehe not my paper but still feels good man

I have a new IO/metrics paper out, with @borusyak, Kevin Chen, and @lihua_lei_stat. We have some new results on the nonparametric identification of demand counterfactuals with recentered IV We still doing tweet threads here? Ok, here goes...

@KarlTowns wearing your jersey while I protest, thank you KAT

What is happening in the Twin Cities and the Great North Star State is heartbreaking to witness. These events have cost lives and shaken families — and we must call for accountability, transparency, and protections for all people. This moment demands that we reflect honestly on what our values truly are. My thoughts, prayers, and deepest condolences are with the families of Renée Good and Alex Pretti. I stand with the people of Minnesota.

@gsatkowiak That’s the Great Hall at Saint John’s in the third photo, right? Super cool work!

art not made with ai

New @cepr_org Discussion Paper out with @CarolineFohlin! We study market behavior throughout the infamous Teapot Dome Scandal in the 1920s, when insider trading was still legal in the United States. A short🧵 (1/4)

Hi all, please spread the word and we hope everyone can make good use of this new data drop: CMFdata.org The full surviving establishment-level Census of Manufactures manuscripts from 1850, 1860, 1870, and 1880!

I love Halloween because it means I can wear the jacket my girlfriend made me into the office :)

Being doxxed as a member of the Ku Klux Klan had "no statistically significant effect on any outcomes of interest". Can't wait to read this JMP

@gp_mihalache I’m a Chugh fan, and the slides are online!

bunching identification scheme goes brr

@pagingkaren This is a first-author level nap right here

@NathanEcon promote him to coauthor immediately

Can someone give me tips on RA management? He is just too sleepy to learn how to code

@LDJEconomics His intelligence is artificial because he keeps snacking on artificial plants

@NathanEcon CatGPT

bootstrapping still feels like magic, no matter how many times I do it

We just posted solutions to exercises of my book ``Open Economy Macroeconomics'' @PrincetonUPress

@abdurrahmanregi While the VRAM constraint is important when running computation with a lot of data, that wasn’t the reason my code was taking a long time. The complexity of the estimation scheme was the challenge, and the GPU allowed for fast parallelization of that.

@NathanEcon congrats! but you don't get hit by VRAM (assuming GPU provided by nvidia/amd/equivalent). or this is done in university server not in personal PC where VRAM constraint is more prominent?

turns out code that takes 3 days to run can be refactored into computation on a GPU and take 5 minutes and now i'm never going to shut up about this

@gp_mihalache This is Julia using the Flux package with ADAM! It plays super well with GPUs. I could have threaded my CPU instead, but it was substantially slower (hours as opposed to minutes, still a major upgrade)