@jasonfurman

Professor @Harvard, Director @HarvardBizGov, Senior Fellow @PIIE, teaches Ec10, words @nytopinion, proprietor https://xitter.cf/t.co/4eOTjrd0x4, was Chair of Pres Obama's CEA.

Cambridge, MA
Joined April 2011
Thousands signed up for Econvitals! But some tried & failed. If you're one: 1. Try again, an overactive anti-bot filter was rejecting some, it's been fixed 2. Type correctly. (e.g., the @imf.orgp & @gmaol.com who tried) 3. Check spam for confirmation econvitals.org/subscribe/
8
2
2
47
11,393
Jason Furman retweeted
Since 2019, consumer confidence has fallen more in the United States than in any other advanced economy the OECD tracks. That has coincided with the U.S. economy generally outperforming most of its peers.
2
8
1
23
4,824
Jason Furman retweeted
The week ahead in U.S. economic data: Tuesday: Durable Goods Orders Wednesday: FOMC decision Thursday: PCE, GDP (Advance), Personal Income & Outlays Friday: Employment Cost Index Prior readings and forecasts in the table.
2
1
11
4,979
I'm thrilled to be launching Econvitals today. It's been the main way in which I have been tracking and understanding the economy over the last several months. And I decided I might as well share it with all of you--for free. Has a combo of website, email lists & social media.
4
3,938
Jason Furman retweeted
A college student reacts to some of what is written by Econvitals.
5
3
98
20,279
A way to summarize and aggregate the latest data flow. Two dimensions: (1) How is the data relative to what happened in the past? How is it compared to what was expected the day before release? (2) Activity and inflation.
The Econvitals composites, week ending October 23: activity is 0.7 standard deviations below its history since 2013 and 0.2 below forecasts. Inflation is 2.1 above its history and 0.2 above forecasts.
6
7
1
41
12,070
The latest readings mean that inflation is very high compared to historical performance but has basically been at forecast. And activity is a bit below historical performance but also basically at forecast. Historical window is since 2013 (as far back as my forecast data goes).
3
1
7
4,363
It is truly an amazing time to build something new. I sort of realized this when I left AV to start my own thing in June. But because I’d used AI so little at that point I completely underestimated how much it approximates having a full team. What a time to be alive.
I asked Claude how much it would have taken to develop Econvitals pre-coding agents. It's answer: a team of about 50 people working for three years with a budget of $20 million. Is this correct or too high? I'm not sure. econvitals.org/about/before-…
2
6
1
71
14,489
A college student reacts to some of what is written by Econvitals.
5
3
98
20,279
I agree the writing is a bit tortured (I hope LLMs improve over time, sometimes they feel like they're getting worse). But the range of evidence (from macrodata to CPS microdata), reasoning (diff-in-diff style), and range of issues and data covered feels outstanding to me.
1
6
4,357
The Econvitals post is much better than what you would get from a chatbot because Blogwright integrates data sets, an architecture with agents including planners, researchers, writers and fact checkers, etc. But my guess is simple chatbot better than Blogwright in 1-2 years.
8
3,636
I mostly agree with you @Claudia_Sahm. I expect colleagues to continue to have research assistants even if only for mentorship than because they need the help. Companies to have junior staff not because they need their current work but to train them for future roles.
Furman picking AI over CEA staff deserves scrutiny. His staff went on to found the Yale Budget Lab, run Treasury economic policy and a Fed institute, serve in later admins, and publish research on trade, environment, and taxes. That's what "20 staff" become. No model does that.
19
14
7
175
43,450
Suppose a grantee for Alzheimers research could do 2X as much without having to pay RAs. Should they do that? Get more research today but possibly at the expense of better future researchers. And is the grant you got about Alzheimers or to support training? Hard questions.
1
21
5,253
And one last point: IF I'm correct (and you may disagree), it is remarkable that a machine can do what twenty highly, highly trained humans could do. I find myself humbled each and every day by them, a feeling I both hate and love in equal measure.
4
3
28
4,436
Jason Furman retweeted
In fact, a human controlled future is worth a cost to long-term efficiency as well.
I agree that if we want a human controlled future, we need to hire and train junior employees even at a cost to short term efficiency.
16
25
3
246
17,051
I asked Claude how much it would have taken to develop Econvitals pre-coding agents. It's answer: a team of about 50 people working for three years with a budget of $20 million. Is this correct or too high? I'm not sure. econvitals.org/about/before-…
34
20
10
210
66,380
There is something exciting and liberating about this. There is also something inefficient about it. And extremely disruptive as well.
2
1
20
4,876
BTW, more about Econvitals. xitter.cf/jasonfurman/status/211…
I'm thrilled to be launching Econvitals today. It's been the main way in which I have been tracking and understanding the economy over the last several months. And I decided I might as well share it with all of you--for free. Has a combo of website, email lists & social media.
2
20
5,274