Matt Blackwell
@mattblackwell.bsky.social
data, causal inference, experiments, politics
https://mattblackwell.org
https://mattblackwell.org
Pinned
Reminder that I have a textbook on an introduction to mathematical statistics and regression, designed for first year PhD students in poli sci, but maybe useful to others. Let me know if you use it and have feedback!
mattblackwell.github.io/gov2002-book/
mattblackwell.github.io/gov2002-book/
data engineering is my passion
February 14, 2026 at 6:14 PM
data engineering is my passion
Say what you will about subjective Bayesianism but at least it’s an ethos.
February 14, 2026 at 6:11 PM
Say what you will about subjective Bayesianism but at least it’s an ethos.
The way I think about this is that the distribution of effects is approximately sparse (most effects are very weak but nonzero). Journals/researchers are interested in effects that are surprising given that prior distribution.
I have a new paper. We look at ~all stats articles in political science post-2010 & show that 94% have abstracts that claim to reject a null. Only 2% present only null results. This is hard to explain unless the research process has a filter that only lets rejections through.
February 14, 2026 at 4:29 PM
The way I think about this is that the distribution of effects is approximately sparse (most effects are very weak but nonzero). Journals/researchers are interested in effects that are surprising given that prior distribution.
An eternal story: there’s a lot of money/clout/influence to be made in scaring people into listening to you.
“Doom is on the horizon, but, lucky for you, I have the answers if you’ll just like and subscribe.”
“Doom is on the horizon, but, lucky for you, I have the answers if you’ll just like and subscribe.”
February 14, 2026 at 2:00 AM
An eternal story: there’s a lot of money/clout/influence to be made in scaring people into listening to you.
“Doom is on the horizon, but, lucky for you, I have the answers if you’ll just like and subscribe.”
“Doom is on the horizon, but, lucky for you, I have the answers if you’ll just like and subscribe.”
Really curious about this 3-4hr tasks that in the METR database that do worse than the longer ones.
February 13, 2026 at 11:43 PM
Really curious about this 3-4hr tasks that in the METR database that do worse than the longer ones.
you really should deal in first principles. Does X(X’X)^{-1}X’y constitute “thinking”?
February 12, 2026 at 1:31 AM
you really should deal in first principles. Does X(X’X)^{-1}X’y constitute “thinking”?
On a lark, tested Opus 4.6 on a common academic task: take a 10k word article and shorten it to 6k words for submission to a new journal. I told it to use the command line tool texcount to count words given the known inability to count words. It failed in a fairly funny way…
February 12, 2026 at 12:24 AM
On a lark, tested Opus 4.6 on a common academic task: take a 10k word article and shorten it to 6k words for submission to a new journal. I told it to use the command line tool texcount to count words given the known inability to count words. It failed in a fairly funny way…
Watching Claude code struggle with word count for a tex file. LLMs, they’re just like us
February 11, 2026 at 3:01 PM
Watching Claude code struggle with word count for a tex file. LLMs, they’re just like us
One of the biggest wastes of academic productivity is the variation in page/word limits across journals in the same/related fields.
February 9, 2026 at 6:17 PM
One of the biggest wastes of academic productivity is the variation in page/word limits across journals in the same/related fields.
I’ve seen takes you wouldn’t believe
February 8, 2026 at 3:18 PM
I’ve seen takes you wouldn’t believe
The python notebook dependence is one of my least favorite parts of the move to industry. Writing code that gets saved as json instead of plain text, absolutely disgusting.
February 6, 2026 at 4:41 PM
The python notebook dependence is one of my least favorite parts of the move to industry. Writing code that gets saved as json instead of plain text, absolutely disgusting.
AI is going to let academics write so many more papers (derogatory)
February 6, 2026 at 12:01 PM
AI is going to let academics write so many more papers (derogatory)
Getting acknowledgment-zoned (feedback ignored, added to \thanks{})
February 5, 2026 at 2:51 PM
Getting acknowledgment-zoned (feedback ignored, added to \thanks{})
Looks like a cool paper but this title, man. Fields other than economics do formal modeling!
February 5, 2026 at 2:00 PM
Looks like a cool paper but this title, man. Fields other than economics do formal modeling!
Two sentence horror story
February 5, 2026 at 1:49 PM
Two sentence horror story
It’s probably not AGI if you have to rely on the “skill issue” arguments to explain its behavior
February 4, 2026 at 2:03 PM
It’s probably not AGI if you have to rely on the “skill issue” arguments to explain its behavior
One oddly moralistic part of my parenting philosophy is that I try very hard to deny my children the knowledge that food delivery exists as a concept.
Ma'am you are spending 1/3 of your post-tax income on doordash
February 2, 2026 at 2:05 AM
One oddly moralistic part of my parenting philosophy is that I try very hard to deny my children the knowledge that food delivery exists as a concept.
Funny that WashU and Vanderbilt don’t make the Manhattan Institute’s Top 20 colleges even after their presidents spent the last year and a half trying to woo conservatives
February 1, 2026 at 12:57 AM
Funny that WashU and Vanderbilt don’t make the Manhattan Institute’s Top 20 colleges even after their presidents spent the last year and a half trying to woo conservatives
Ross Douthat and other prominent posters are kinda losing it on twitter about this and all I can think of the 1896 short film Train Pulling Into the Station which had this reaction
January 31, 2026 at 12:20 AM
Ross Douthat and other prominent posters are kinda losing it on twitter about this and all I can think of the 1896 short film Train Pulling Into the Station which had this reaction
You’re absolutely right—it does look like our collider has produced a black hole that is expanding rapidly.
Here are some fun ways to pass the time before you hit the event horizon:
🍿 Put on Interstellar
🍝 Read up on spaghettification
🤗 Hold your loved ones tight
Here are some fun ways to pass the time before you hit the event horizon:
🍿 Put on Interstellar
🍝 Read up on spaghettification
🤗 Hold your loved ones tight
January 28, 2026 at 1:36 AM
You’re absolutely right—it does look like our collider has produced a black hole that is expanding rapidly.
Here are some fun ways to pass the time before you hit the event horizon:
🍿 Put on Interstellar
🍝 Read up on spaghettification
🤗 Hold your loved ones tight
Here are some fun ways to pass the time before you hit the event horizon:
🍿 Put on Interstellar
🍝 Read up on spaghettification
🤗 Hold your loved ones tight
AI techno-futurism is wonderful until I remember the phrase “where are the flying cars I was promised”
January 26, 2026 at 3:28 AM
AI techno-futurism is wonderful until I remember the phrase “where are the flying cars I was promised”
my coping mechanism for the news cycle is learning tmux
January 23, 2026 at 3:41 PM
my coping mechanism for the news cycle is learning tmux
Interesting tidbit from the lates YouGov poll: only ~50% of 2024 Trump supporters identify as MAGA.
January 21, 2026 at 1:30 AM
Interesting tidbit from the lates YouGov poll: only ~50% of 2024 Trump supporters identify as MAGA.
One thing I’m always a little confused about “adapt or die” AI narratives: if AI progresses as fast as “fast takeoff” people think, won’t your specialized knowledge of stuff like Ralph be obsolete? Won’t an AGI just be able to implement Ralph w/o you knowing anything about it?
January 20, 2026 at 11:34 PM
One thing I’m always a little confused about “adapt or die” AI narratives: if AI progresses as fast as “fast takeoff” people think, won’t your specialized knowledge of stuff like Ralph be obsolete? Won’t an AGI just be able to implement Ralph w/o you knowing anything about it?
If you are wary of letting LLMs write your statistical code, you can ask one of the CLI tools (claude code, gemini cli, codex) to be a code reviewer for an empirical project that you are working on. I told it to pretend to be a reviewer at a top-5 econ journal for extra pettiness
January 20, 2026 at 12:03 AM
If you are wary of letting LLMs write your statistical code, you can ask one of the CLI tools (claude code, gemini cli, codex) to be a code reviewer for an empirical project that you are working on. I told it to pretend to be a reviewer at a top-5 econ journal for extra pettiness