iiinova — where your digital self keeps evolving

Reposted by ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️

Rebecca Stephens Falcasantos @rsfalcasantos.bsky.social · 07/08/2026

Our department is hiring. TT position in Hebrew Bible and Ancient Judaism. I’ve found great colleagues and students here. Please share! apply.interfolio.com/189845?fbcli...

apply.interfolio.com

Apply - Interfolio {{$ctrl.$state.data.pageTitle}} - Apply - Interfolio

02931

Reposted by ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️

John Mullahy @johnmullahy.bsky.social · 07/08/2026

Fifty years ago this month the JPE @jpolecon.bsky.social published this iconic special issue on labor and family economics. www.jstor.org/stable/i331901

1153

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 08/08/2026

I cited several of those papers I believe in my job market paper

010

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 05/08/2026

His wife's maiden name is Usha Bala Chilukuri. How do people put up with this bigotry?

2174

Reposted by ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️

Paolo Brunori @paolobrunori.bsky.social · 04/08/2026

Call for Papers for the Equality of Opportunity and Social Mobility Policy Workshop, to be held on 12-13 November 2026 at the London School of Economics. Submission deadline: 21 August 2026 More information: www.louissirugue.com/eosm

022

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 03/08/2026

Another option that has not attracted the attention of applied researchers is model averaging (aka "stein-like estimators"). Instead of pre-testing and choosing 1 instrument or 1 estimator, average across all candidates. I wrote about this here. dlm-econometrics.blogspot.com/2020/04/to-c...

dlm-econometrics.blogspot.com

110

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 03/08/2026

say fixed and random effects or OLS and IV. What is the solution? Surely if you try to publish a paper with a weak first-stage and tell the referees "at least I did not pre-test" it will not go well. But, this is a referee issue. Appeasing misinformed referees is not a justification for bad science.

110

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 03/08/2026

and (ii) searching over many candidate instruments and using only the strongest. Again, not specific to IV. Everyone's darling, diff-in-diff, likely performs similarly poorly if one only retains applications with "no pre-trends". Other situations such as using a Hausman test to choose between

120

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 03/08/2026

A while back, when blogs were still a thing and I had the mental bandwidth of more than a grain of rice, I wrote about this. A simple simulation shows how truly awful (i) putting weak IV papers in the "file drawer" and only keeping papers that "work" dlm-econometrics.blogspot.com/2021/05/cant...

dlm-econometrics.blogspot.com

Can't Hack It!

As the academic year from hell reaches completion, the sun emerges, and the ability to go on vacation resumes, we should all have a little s...

110

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 03/08/2026

There is doubt that this is standard practice. And for those who are thinking "this is why i hate IV", the issue is not IV-specific. "Pre-testing" and then going with the estimates that pass the pre-test (in this, a "strong" instrument) completely changes the statistical properties of the estimator.

120

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 03/08/2026

The current issues of @aeajournals.bsky.social J Eco Perspectives has a nice primer on weak instruments and IV more generally. One of the "Don'ts" for applied researchers is copied below: don't search over a pool of instruments for a "strong" one and then go with that IV estimate.

1227

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 01/08/2026

Don’t sell yourself, um, short. You bring down the height. 😂

020

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 31/07/2026

It’s the worst part of using partial identification. It’s been a while but there are a few other approaches but hard to know exactly what to do as an applied person.

110

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 31/07/2026

More like this, but with 4 of us. We need a short person for Alfredo!

020

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 20/06/2026

They shouldn’t, but they should care about wasting $15M on a pool that’s falling apart

040

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 04/06/2026

Ahhhh those were good times. Properties of estimators given the assumptions one makes about model being estimated. Those assumptions can be shown in a DAG or they can be expressed in terms of attributes of the things in the estimating equation. I prefer the latter.

100

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 04/06/2026

@p-hunermund.com is the DAG champion. I also get the impression that some times at least the DAG is taken as the true model and the need to justify the assumptions embedded is not seen as necessary. That probably just as equally to people who start by writing down an estimating equation.

120

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 04/06/2026

I am totally with you. I find it much more intuitive to think about the econometric properties of an estimator for a given model. If a DAG suggests a problem, like conditioning on a collider, there has to be an econometric reason as well. That econometric reason is easy and intuitive to me.

130

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 20/05/2026

Not of limited use per se, but used properly. There is a lot of “devil is in the details” in econometrics and — as Marc discusses in another comment — economics doesn’t like to publish these articles as they are often in the wasteland that lies between applied researchers and serious econometricians

100

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 12/05/2026

I have no idea if this is true or not, but … data can be faulty?! 😱 who knew? 🤷‍♂️

041

Reposted by ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️

John Mullahy @johnmullahy.bsky.social · 12/05/2026

ABAA

111

Reposted by ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️

Will Wheeler @willwheels1.bsky.social · 12/05/2026

My office (NCEE) is hiring an RA/pre-doc. This is an excellent opportunity for students interested in environmental economics. Term is for one year with possibility of renewal. Bachelor's or Master's required by 8/17. #econ_ra www.zintellect.com/Opportunity/...

zintellect.com

Zintellect - U.S. Environmental Protection Agency (EPA) - EPA Fellowship in Environmental Economics - EPA-OP-NCEE-2026-0001

*Applications may be reviewed on a rolling-basis and this posting could close before the deadline. EPA Office/Lab and Location: A research training opportu

022

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 11/05/2026

you can get away with it because no one really knows what entropy is

220

Reposted by ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️

John Mullahy @johnmullahy.bsky.social · 10/05/2026

Happy Mother's Day.

141

ᴅᴀɴɪᴇʟ ᴍɪʟʟɪᴍᴇᴛ ✡️☮️❤️ @dlmillimet.bsky.social · 07/05/2026

Call me IV (because I’m biased), but I really like this paper, had fun working with Marc, think it is very simple but important for applied people, and provides a simple (partial) solution to the problem. We both hope researchers will take note. PS This paper has nothing to do with diff-in-diff.

0140