Arpit Gupta

Was there in person but sadly had to leave before testimony

Under completely inelastic housing supply, yes. But I think more realistically improved mortgage access also helps push out supply even under current regulations.

Simple way to see this is lower rates -> more building post 2020, and higher rates -> less building. This is a margin.

This is basically the reason I think why mortgage diversity dried up. We used to have a bunch of options: price level mortgages, mortgages which extended/contracted the payment length (while keeping payments the same), but Fannie Freddie contracted the space

Note you don’t have to offer subsidies for this product; it was produced by the private market without subsidy in the US (and various types overseas)

This is a really good discussion of “applied” AI in a non-coding/SWE space.

Sure, but I think people actually underestimate the elasticity for single family homes financed from traditional mortgages.

Yes, it's hard to build in constrained metro areas. But conventional estimates of supply elasticities are still reasonable.

I did a bit of searching around, and all of this was a massive research area in the 70s-80s. People thought about (and people did originate) all sorts of mortgages, including this "PLAM" option with fixed real payment

Biggest downside is hardship of escalating payments, generating possible mismatch

2. When Your AI is Lying To You Financial Document Intelligence, Hallucinations, and how to Ground AI in Reality

Week two of my AI in Finance course:

• Financial document intelligence
• The jagged frontier of AI capabilities
• Why your AI is lying to you, will we ever get to truth, and how to use RAG to deal with it

arpitrage.substack.com/p/2-when-you...

We used to have these (tilt, or graduated payment mortgages)

Thanks! Timing of this course for me is not great but will try to keep up with updates. One of the motivations there for the open structure was your metrics class.

I'm teaching a new course at Stern on AI in Finance and opening it up!

Syllabus/slides on Github: github.com/arpitrage/ai...

Weekly summaries on Substack
arpitrage.substack.com/p/1-three-ru...

First post is on Amdahl's Law, Jevons' Paradox, and why finance was slow to learn the bitter lesson

Joke’s on you I don’t like school either

This one's for you @arpitrage.bsky.social

Lot of good posts on Twitter still

Huh? All we talk about on Twitter now is AI.

Good: public officials are often underpaid, but there should be a reasonable process for processing raises

In the UK, Australia, and India: do you think they tell promising but not yet there junior academics who need higher impact papers “Go on, hit a six paper”

Economists should start doing economics

In the paper too, but you can see it here

this is consistent w what I’ve been observing in my and friends’ careers post-pandemic; if you want to continue working remotely, it is unintuitively *more stable* to favor startups (doesn’t have to be super-early-stage! love a series B company myself) than try to hang on someplace that wants RTO.

Interesting study here showing that availability of remote work helps startups get top employees from big companies, helping recruiting at startups but hurting retention at big companies. Unclear if return-to-office (RTO) polices help or make things worse for big companies (my guess, worse).

We have some evidence of that — remote workers tend to actually be lower productivity in their prior job, but are higher productivity at the remote firm, which is at least consistent with a good match

1/ When I worked at AWS during the pandemic, internally, we were told remote work was here to stay, Amazon was now a remote work first company, and they were seeing huge productivity gains. This was all in support of the "Earth’s Best Employer" leadership value.

Then RTO hit out of no where 🧵

👀 Super-interesting paper on remote work, productivity from Arpit and co-authors.

Also has some important lessons for big cities and smaller places👇

That doesn’t fully address your question though; but I tend to agree firms will experiment and sort themselves

The benefit we have in our design is while we measure GitHub productivity, we have variation across firms. Ie some firms have a lot of remote friendly workers; others do not, but both hire some coders.

Remote Work's Impact on Productivity Why Startups Thrive while Big Companies Struggle with Remote Work

Good read on remote work from @arpitrage.bsky.social

arpitrage.substack.com/p/remote-wor...

• Remote work helps to balance the geography of innovation. If startups can hire remote workers in a range of locations, we might see a broader dispersal of innovation and other activity across metros, rather than being concentrated in a handful of cities.

Feedback welcome!

These results suggest:
• Remote work is challenging to pull off, with tradeoffs between labor market access and productivity. These tradeoffs are best managed by startups who gain a competitive edge against large firms (who are instead going RTO)

This channel of increased scale accounts for about half of the impact of remote work on productivity. The benefits are seen in the productivity rate of new hires, and also seem to spillover to other workers because existing employees also benefit when new workers join.