Max Kagan

Besides, scholars furious about reviewers uploading manuscripts to AI, but fine with Elsevier paywalling them for $50 so 12 people can read them, have a very interesting definition of "consent."

Weren't we supposed to be using taxpayer money to make publicly funded research actually read and used?

to be fair though, there should always have been more integration of (a) campaign finance and lobbying literature and (b) studies of US and developing / less institutionalized / more clientalist contexts

I think this conflates the money in politics field (can money sway an individual election ➡️probably not) with the lobbying literature (can licit and/or illicit exchange buy favorabel policy outcomes ➡️not sure there's ever been much doubt there)

Meanwhile, Wiley journals do not allow peer reviewers to use AI to evaluate manuscripts.

Just made a manuscript/PAP feedback GPT built from 150+ previous peer reviews - identifies problems and provides actionable feedback on the points that I raise most frequently
chatgpt.com/g/g-68af4d19...

Anyone can use - try it out! (Editors/authors: feel free to cut me out of the loop 😉.)

Yeah—I don’t think any of us can know. Hopefully critical thinking is a lowest common denominator that is valuable regardless of what future emerges. And curiosity feels underrated in discovering how to work in this AI world and in general. But not sure we know how to teach either of these!

You can start experimenting with the $20/mo tier, but once you start using it you will quickly find it’s worth it to pay $100 or $200 month so as not to get quickly rate limited. (I say this as a postdoc with a reasonable—but not crazy—research budget.)

When I pitch academics on my paper on nulls one common and understandable reaction is "but they're probably noisy and thus uninformative nulls." This is true, but it misses the key realization that WE PUBLISH THE RESULT WHEN THE NOISY TEST IS P<0.05.

I also think the potential for replicating papers with Claude Code can be a huge teaching resource for those so inclined. You could imagine building a more practitioner-oriented class around having Claude Code interactively dissect replications to figure out what is good research

i.e., it might not be wise to have an undergrad try to "vibe code" a DiD paper, but one could imagine a senior seminar project where students collect original descriptive data at scale

The vast majority of undergrads were never going to be RAs, though. For your modal undergrad (i.e., someone who is vaguely interested in law school), this could actually perhaps reduce the gap?

Also see Jon's comment, which expresses my feelings about this really well. If you look at the paper and think about what *kinds* of tasks RAs would be asked to do for this, you'll see there are much better uses of their time (for both learning and productivity objectives). bsky.app/profile/jonm...

There are solutions to the problem of null results and we know them.

One of them is the widespread introduction of Registered Reports across journals.

It is not that we lack solutions but that we lack the will to change things (and accept a potential decline in the IF that we all love so much)

It must be very hard to publish null results Publication practices in the social sciences act as a filter that favors statistically significant results over null findings. While the problem of selection on significance (SoS) is well-known in theory, it has been difficult to measure its scope empirically, and it has been challenging to determine how selection varies across contexts. In this article, we use large language models to extract granular and validated data on about 100,000 articles published in over 150 political science journals from 2010 to 2024. We show that fewer than 2% of articles that rely on statistical methods report null-only findings in their abstracts, while over 90% of papers highlight significant results. To put these findings in perspective, we develop and calibrate a simple model of publication bias. Across a range of plausible assumptions, we find that statistically significant results are estimated to be one to two orders of magnitude more likely to enter the published record than null results. Leveraging metadata extracted from individual articles, we show that the pattern of strong SoS holds across subfields, journals, methods, and time periods. However, a few factors such as pre-registration and randomized experiments correlate with greater acceptance of null results. We conclude by discussing implications for the field and the potential of our new dataset for investigating other questions about political science.

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.

Check out the new paper! And please please please send us both gut reactions and harsh criticism.

This short thread is worth a read, but this stands out to me.

I've found that null results often elicit a "yes, but what about..." response. Yes, we don't find any effect of x on y, but obviously (despite previous assertions by others) the real effect of x is on z.

I mean "just talk to it" literally! I start every day with a rambling conversation to Claude about what's on my mind: random ideas that span the gamut from "figure out how to run a new robustness check" to "can't forget to pick up milk later," and it makes sense of all of it

It’s the same with Claude Code. You just have to use it, you just have to trust that in the limit — and trust me, you reach that long run limiting point within no time, seriously — it will figure you out. That’s the thing — it won’t be you figuring out Claude Code. It’ll be Claude Code figuring you out! So you just have to do it. You don’t start with starter kits. You start with directories. I mean it! You start with this: Please look around this directory and tell me what the hell is going on in here. It looks like a hoarder’s nest! Please help me make this deck of slides for my class! I’m overwhelmed with work. Use beamer, and install it if it’s not installed. I don’t even know how to use that frankly. Do the whole thing for me. I am anxious about the class, and I am anxious about the research project, and I am anxious about you. I need help. So please make a beautiful, beautiful deck based on this chapter, on this thing we’re doing in class, and here — here’s last year’s exams, here’s last year’s homework. Here’s a couple of papers. Just please read them, and come up with a beautiful deck. Make my life easier, not harder. I am living on my budget constraint and if you can help me, I will really appreciate it.

@causalinf.bsky.social nails the experience of using Claude Code and how it is so different from what has come before. People want to know "how do I use the AI" and the answer is simple: "just talk to it!"

On the "does ideological moderation help candidates win debate?":

- candidates' ideology reflect strategic choices (endogeniety)
- perceptions of candidate ideology are related to candidate quality (endogeneity)
- favorable race dynamics attract higher-quality candidates (endogeneity)

🤔🤔🤔

🧵 New version of our paper (@bcegerod.bsky.social) is finally online: "How Many is Enough? Sample Size in Staggered Difference-in-Differences Designs"
We show that even well-identified DiD studies are often underpowered; sample sizes needed are surprisingly large
Paper: osf.io/preprints/os... 1/6

Findings aside, it is weird how journalists switch back and forth between percentage points and percentages in the same article

This is true and an important note. But it's not just tuition. Yes they provide (generous) financial aid, but it can still be a very weird, lonely, and isolating experience to be in a school where the median student is >90 percentile family income

Explore How Income Influences Attendance at 139 Top Colleges (Published 2023) Unlike many elite colleges, most flagship public colleges are “pretty fair in who gets in.”

In sum: liberal arts colleges really do offer amazing teaching and opportunity. But the fact that these schools admit more students from the 10% of the income distribution than from the remaining 90% cannot be absent from the conversation.

www.nytimes.com/interactive/...

To be clear, I loved Midd. But one of my most abiding memories was someone who wrote an editor to the school newspaper during their senior year asking the career center to stop reminding him to attend job fairs because it was distracting from his enjoyment of his education.

But this exists alongside immense privilege. It's no accident that these places skew so wealthy. The community college transfer and first gen students I taught at UCB often to had to balance school with working, supporting family, and finding a well-paid job to pay back loans.

Thanks for sharing. I loved my experience at Middlebury. But after having been a TA at a very different but no less elite school (UC Berkeley), I have mixed feelings. SLACs provided me with a relaxed way to really explore new ideas, courses, etc. A true liberal arts education.

Things are grim. But in more frivolous news...

@jamesbrandecon.bsky.social and I have been chipping away at `dbreg`, a 📦 for running big regression models on database backends. For the right kinds of problems, the speed-ups are near magical.

Website: grantmcdermott.com/dbreg/

#rstats

[1/2]

Call for Proposals: Data Collection for Replication+Novel Political Science Survey Experiments Alexander Coppock and Mary McGrath January 27, 2026 We invite proposals for a survey experiment replication+novel design competition. Se- lected replication+novel design survey experiments will be conducted on large samples of American respondents, quota sampled to match U.S. Census margins and filtered for quality and attention by the survey sample provider Rep Data (repdata.com). Each proposal consists of two parts: (1) a replication study of an existing, previously published survey experiment, and (2) a novel experimental design on a topic of the authors’ choosing. The replication studies and reanalyses of the existing studies will be combined into a meta-paper to be co-authored by all authors of accepted proposals along with the princi- pal investigators (Coppock and McGrath). As a condition for acceptance, authors commit to sharing the data and producing a write-up of the findings from their novel design for submission to a scholarly journal, and public posting of a working paper pre-publication.

🎺 Call for proposals 🎺

1️⃣ replicate an existing experiment
2️⃣ run a novel experiment

on repdata.com

3️⃣ coauthor with Mary McGrath and me to meta-analyze the replications and existing studies
4️⃣ publish your study

details: alexandercoppock.com/replication_...
applications open Feb 1

please repost!

Why the Tech World Thinks the American Dream Is Dying Silicon Valley fears this is the last chance to amass generational wealth before AI makes money worthless.

SV needs to pick up Polanyi (more of a slog than James C. Scott, but worth it). AI will be disruptive but the idea that it will inescapably accelerate inequality only follows if you assume societal institutions are not themselves likely to be disrupted.

www.wsj.com/tech/ai/why-...

Which model(s) and setup are you using? I don't think most out-of-the-box chatbots are going to do a great job with this sort of task (context window is too small)