Adam Zelizer

Hmm must be confusing fans with leadership of the athletic department.

Look, as a born and raised Tennessean, we have to give credit where credit is due. It nailed Kentucky and Alabama, whose basketball and football fans, respectively, must have provided plenty of training data.

This only holds if think presidential approval ratings measure anything meaningful in American politics today. Is focusing so much energy on them in both academic and journalistic work helpful for understanding the world?

(I don't think so.)

The Midwestern Congressman Whom Washington Should Remember An appreciation of the late Rep. Lee Hamilton by a veteran congressional reporter.

Read Politico's David Rogers on Lee Hamilton. www.politico.com/news/magazin...

I have some survey rcts now in the field. You’re right - on ICE they do not, as of today, follow elite cues. Appears they know what they like and will turn on elites who don’t take their preferred position. On less salient (but still partisan) issues, I see huge follow the leader effects.

Doesn’t quite demonstrate good causal reasoning. Is there a fundamental constraint in AI understanding causal inference, or is it simply that they are trained on databases of old papers pre-credibility revolution?

Huh, survey respondents differ from nonrespondents in unobservable and unpredictable ways.

To be clear I'm referring to OP, not Dan.

Professors publicly shaming a phd student at another university is bullshit. Write to their advisor - even to their IRB if you think there's a real issue. Don't post about it on Bluesky (or write about it in your APSA presidential address, but that's another story).

And what publicly available data means! You can't complain about data being linked to you when that information is public. JFC you don't have to consent to someone collecting public information about you.

All other issues aside, this complaint reads like it's from a guy who's had it up to here with his own IRB bullshit.

post a perfect album from the 90s that isn't nirvana, pearl jam, soundgarden, or Alice In Chains. Abiding by the letter, if not the spirit, of the prompt.

Revisiting a Natural Experiment: Do Legislators with Daughters Vote More Liberally on Women’s Issues? | Journal of Political Economy Microeconomics: Vol 1, No 3 The pioneering work of Washington in 2008 shows that legislators with daughters cast more liberal roll call votes on women’s issues. Costa and coauthors in 2019 find that this pattern subsides in more...

Doesn’t hold up.

www.journals.uchicago.edu/doi/abs/10.1...

That’s a good question. I would think the main reason we’d be interested in this result would be basically academic - to better understand the psychology / decision-making of voters.

Subsetting the data based on # of kids is another version of using # of kids as a covariate. It's still a bad control. The right thing to do is only estimate an effect of the first child's gender.

That's relatively small potatoes - how do you get separate estimated effects for boys & girls?

Also, there is a widespread parental preference for a mix of genders, so # of kids, beyond 2, is a bad control. One could consider only gender of first-born (or first and second). The more conservative approach is safer if there is any gender preference, which could vary across families.

Something is wrong. This analysis identifies the differential effect of having a boy vs. girl. There is no credible ID strategy for effect of having a boy AND effect of having a girl using biological randomness of gender.

One thing I’ve learned from my colleagues - theorists are very smart and seemingly always able to come up with a credible alternative to the causal story Im telling them.

Hi Kenny! His cue-taking paper is very much a neat natural experiment.

I find Lee’s explanation compelling - parties started to approach roll calls more strategically due to electoral competitiveness (and electronic voting), which caused the appearance of both gridlock and polarization. I believe both causal stories and have difficulty distinguishing them empirically.

In my experience, when that happens - a credible, appealing causal claim runs into contradictory, weak evidence - audiences suddenly turn into a collection of the most rabid randomistas, finding all the reasons the empirics should be questioned. When the same evidence confirms the theory, silence.

I think we agree there are compelling causal claims based on theory (ie polarization and gridlock). But does unidentified empirical work, like polarization and gridlock both increasing in Congress, make the claim more compelling? Would contradictory evidence undermine the theoretical mechanism?

Wow, this is how you do it. Does anyone read this and think any less of the original authors? Or does your esteem for them go up?

In general, as I don’t know this case, and to bring the 🔥, I don’t find multiverse analyses compelling. If there are 1000 specifications that could make sense, then there probably is no research design or reason to believe the results.

It doesn’t make them sloppy or bad researchers. I’m sure everyone has a result that is weaker than it looks. There are top researchers who have real howlers of errors or misleading analyses (good luck to anyone trying to publicize those). And we don’t have to protect our friends from criticism.

And just because someone has published something wrong doesn’t mean they have to dig in and fight the replication! We need to normalize saying “maybe the evidence isn’t as strong as presented. I still think it is convincing, but there are, as always, reasons to be skeptical.”

I want to give kudos to @dandekadt.bsky.social and @apsrjournal.bsky.social for doing the thing we always ask: publish critiques or replications; don’t give priority to published work just because it’s published; use editorial discretion; stick your neck out to criticize work you think is wrong.

Am I wrong to read this as "Hey, we spend a bunch of money on agriculture research anyway. Some of it looks at tech, AI, etc, so what the hell, let's make it look like a political win for the administration."