Rachel Leah Childers

A new blog post motivated from this amazing discussion with @akhilrao.bsky.social and @donskerclass.bsky.social

#EconSky #RStats #Stan

jamesblandecon.github.io/posts/2026-0...

Soon enough, people didn't even mention that they had used it, let alone give a proper citation. Frequently the reader was left to guess based on the plots what kind of settings they had used.

A Problem that Stumped Milton Friedman This website presents a set of lectures on quantitative economic modeling, designed and written by Thomas J. Sargent and John Stachurski.

This reminds me, for the case of beta-bernoulli where the choice is A vs B vs continue rather than bet vs continue, QuantEcon has a nice series of lectures with code. Even with conjugacy you need approximate dynamic programming since the mean parameters is a continuous scalar state.

Right, the main motivation for approximate approaches here is that the dynamic program requires taking a posterior as state which is high or infinite dimensional even for a scalar continuous parameters if non-conjugate. This is why the bandit literature is so enormous!

For an econ style interpretation of what's going on here, e-values generalize a Kelly multiplicative growth criterion, while dynamic programming optimizes an additively separable expected utility. This was the dispute that provoked Samuelson's famous one-syllable words only paper

Game-Theoretic Statistics and Safe Anytime-Valid Inference Safe anytime-valid inference (SAVI) provides measures of statistical evidence and certainty—e-processes for testing and confidence sequences for estimation—that remain valid at all stopping times, accommodating continuous monitoring and analysis of accumulating data and optional stopping or continuation for any reason. These measures crucially rely on test martingales, which are nonnegative martingales starting at one. Since a test martingale is the wealth process of a player in a betting game, SAVI centrally employs game-theoretic intuition, language and mathematics. We summarize the SAVI goals and philosophy, and report recent advances in testing composite hypotheses and estimating functionals in nonparametric settings.

A frequentist approach to optional stopping-robust ("anytime-valid") inference is provided by a class of methods typically based on e-values, often with a betting or (super)martingale interpretation.
This Statistical Science intro paper isn't bad; Ramdas also has a book and many papers.

THANKS.
Heterogeneous agents are generally a neutral to good thing to try to understand, and there's plenty of very good research in the area. But reading published work conditions on a collider, inducing a maybe negative correlation with other good things like coherence or empirical plausibility...

I agree that signal noise level may influence how nepotistic this sorting process is and do think that a return to the pre-Samuelson aristocratic or ideological selection mechanism, with some new technical trappings, is a plausible outcome if no new clear technical niche can be found.

I kinda think elitism is basically a conserved quantity in econ, since the real raison d'etre of the field is advising real policy and the content is a sorting mechanism for who gets to be in those few positions. E.g. causal revolution replacing traditional econ logic but leaving networks in place.

Definitely a large part of the effect was a reduction in macro overall, and what remained still had plenty of flaws, with lots of bad work in technical areas not covered by Dynare that could still get by in journals due to difficulty (*cough, het agents*). But I think the field did improve on net.

I will say I knew a lot of old macro guys whose human capital depreciated and were really mad about it. But that was mostly a good equilibrium; publish the same amount but raise the bar. The old kind of work still got done, but publications moved into complements.

I often used to think of the model of Dynare. What used to be a months long expense to tweak a parameter in a DSGE became a weekend project. 90s JME etc had been full of those minor tweak papers, and then they disappeared, as standards were raised for interesting and empirically grounded work.

"I'm in this photo and I don't like it" meme.

To give you a sense of how it feels, if I were trapped in a room and forced to take in messages written in Chinese and send out responses from a codebook all day everyday, I would not appreciate if every discussion of my situation were about consciousness and none at all were about how to get out.

To give you a sense of how it feels, if I were trapped in a room and forced to take in messages written in Chinese and send out responses from a codebook all day everyday, I would not appreciate if every discussion of my situation were about consciousness and none at all were about how to get out.

I love that. I do similarly when teaching policy, & typically sneak in half a lecture on epistemology when teaching data analysis. There is unfortunately a lot of metaphysical work on gender that is completely useless in a discussion of, eg, rates of homelessness, poverty, and sexual assault.

Sorry to be snarky; I think there actually is quite a bit of depth to philosophy of justice worth being explicit about.
Work by credentialed philosophers that purports to be about trans people has however been mostly terrible and irrelevant to material conditions and issues described in this book.

My advice for starting something like this is to be unemployed, which was the case when I started it; interest has slowed so it's much less of a commitment now.
Also, keep a spreadsheet, and do preliminary discussions by email to get to the part where you know enough to help.

Sure, fine, there's philosophy in every leaf of every tree. To be specific, there is no claim that trans people present any philosophical issue distinct from those of other people. The theory of justice in the book is straightforwardly socialist, which you can like or dislike as you please.

Book cover: "The Transgender Issue: Trans Justice is Justice for All" by Shon Faye

This week's reading: Shon Faye's "The Transgender Issue"
If I could recommend cisgender people to read one single book about us, it would probably be this one. It's not personal, it's not prurient, there's no philosophy or metaphysics; it's just facts and policy.
www.goodreads.com/review/show/...

It is deeply funny that there are distinct overlapping communities within economics with a research focus on "inequalities", with completely divorced meanings of that term. There's an opening for work on "Moment inequalities of opportunity".

Maybe because I come from the structural side, my first instinct is at least to sketch out a structural model. In almost all such models I've seen with interesting dynamic policy effects, there is feedback, anticipation, nonlinearity, mean reversion, or another feature that violates DiD assumptions.

Hehehe

"Proper understanding of the answers of the above questions should in most cases make you at best ambivalent about DiD. If you still think you have a DiD problem, expect me to try to help you figure out what else you should do"

Cross-Fitting-Free Debiased Machine Learning with Multiway Dependence This paper develops an asymptotic theory for two-step debiased machine learning (DML) estimators in generalised method of moments (GMM) models with general multiway clustered dependence, without relyi...

Update for self-reference: they changed the title to the more informative "Debiased Machine Learning U-statistics"
No clue if my complaint had any part in that. arxiv.org/abs/2206.05235
Also; a separate new paper avoids cross-fitting entirely. Though the full results are restricted to VC classes. 🙁

Narrative mixed strategies

Free Weekly Econometrics Office Hours Email: rachelleahchilders@gmail.com or Sign up form: https://forms.gle/tSk48w1msZgrCUqX8 Time: Tuesdays 4:00-6:00PM Central European Time (GMT +1) (US Eastern Standard Time 10:00AM-12:00PM) (or by appointment) Location: Zoom Link https://uzh.zoom.us/j/62894466032?pwd=M28RFSrc44buAIJtPdvBEeYAIcClip.1 Who: Anyone. Grad students, researchers, government workers. Private sector is okay but in that case if your question requires work that exceeds the allotted time I may request to negotiate a consulting fee. What I can probably help with: Theory questions. Research design. Modeling. Particular expertise: Time series. Causal inference. Bayes. Structural approaches. Machine learning. Theory: Asymptotics. Statistical learning. Bayes/MCMC. Identification. Decision theory. Semiparametrics. Fields: I know most about macro (DSGE, heterogeneous agents, VARs, etc), can help with finance and applied micro (labor, development, health, etc), and can follow along with statistics problems in other areas. Code: I think in R, can write Julia, and can get by in Python. I am likely to suggest you build a model in Stan. I know Stata but if it’s relevant to your question I suspect you can get better help elsewhere.

For Spring semester, I'm bringing back free weekly open office hours for anyone in the world with Econometrics questions.
Tuesdays 4-6 PM Central European Time (US EST 10AM-12PM) or by appointment; sign up and drop by!
Details and signup at: donskerclass.github.io/OfficeHours....

It is with great disappointment that I was unable, in a review of a long history book, to avoid the use of the word "magisterial". Unfortunately, in this case it was difficult to do otherwise.

This book was really good: I had previously not known much about the British empire and now I know quite a lot. I will likely need to spend some time rearranging my worldview to process the information.
www.goodreads.com/review/show/...

Same! Until I saw this paper, I kind of figured Lean (etc) wasn't going to be useful for me. FWIW, this is a proof of concept and there's still a ton even in those textbooks, let alone contemporary research, not formalized, but it shows you can do it, and with code that reads as fairly natural.

?
This looks like a good survey of its topic, but I'm not following the connection to my question.