The Mixtape with Scott is again. Season 5. Season 5 of the Mixtape with Scott goes to be completely different, and enjoyable, and completely different, and artistic! It’ll be known as The Odd Couple. And it’ll be known as “The Mixtape with Scott (That includes Caitlin Myers)”. It’ll have completely different naming conventions till Caitlin decide one we like! Let me inform you all about it.
I began the podcast round 4 years in the past as a approach of making an oral historical past of economics whereas additionally tracing out the historical past of the credibility revolution by means of Orley Ashenfelter, his college students, and the Industrial Relations Part at Princeton. I tacked on a bunch of different issues too alongside the best way like “the scholars of Gary Becker” and “economist within the tech business”, in addition to any variety of eddies I needed to swim in alongside the best way.
And after 130 interviews, I roughly felt like I had tapped my creativity out. I largely got here to grasp the evolution of causal inference a specific approach, which I wrote up throughout a number of substacks, in addition to added all through my new guide, Causal Inference: the Remix (proofs got here to me at the moment actually). It was very rewarding. Perhaps at some point I’ll write up the interviews as a guide (even Claude Code can not but try this), however for now, I’m simply prepared to maneuver on, as 130 interviews is quite a bit.
However transfer on to what? Properly, that’s what I wish to inform you about now. Right now’s episode is the primary episode in a season I’m calling “The Odd Couple” that includes the sensible economist, Caitlin Myers. And the idea is easy:
Caitlin Myers and me will begin a analysis mission collectively which is just carried out on the podcast. And we’ll use Claude Code to do that mission on the air. Whereas doing it, we’ll discuss and snicker and share our ideas about what we’re doing. Consider Bob Ross speaking whereas he paints bushes. Solely as an alternative of bushes, it’s estimated dosage parameters of abortion clinic closures’ impact on marriage utilizing steady diff-in-diff. And as an alternative of a brush, we’re utilizing Claude Code who’s utilizing R, python and Stata. However apart from these trivial particulars, it’s precisely like Bob Ross, or perhaps the View.
Caitlin Myers is the John G. McCullough Professor of Economics at Middlebury Faculty in stunning Vermont. And he or she is, on the time of this writing, arguably one of many main economists engaged on reproductive coverage in the USA, perhaps the world. She’s been printed quite a bit on the subject for a really very long time, together with this text within the Journal of Political Financial system, our JHR on abortion clinic closures, and quite a few others. You will discover all of it at her slick web site.
She’s additionally been a contributor to the general public good by creating public knowledge repositories. She constructed this dashboard. She is aware of the place each clinic opened and closed and when, going again a long time. She’s meticulously described every related legislation regulating abortion entry. In case you’ve learn a paper within the final ten years about abortion providers, there’s likelihood a design by Caitlin, or knowledge she helped curate and distribute, was one way or the other related to it. Her affect on this house has been huge.
However along with being nice, she’s additionally humorous, considerate, and thinks rather well on her toes. Which is without doubt one of the causes I assumed it will be nice to have her as my analysis accomplice and dialog accomplice on the podcast. As a result of I feel if this idea goes to work, lots of planets should align, and I had been considering for a really very long time that if there was such a sq. peg to suit a sq. gap, it will be her.
I’d say that Caitlin and I are proper at that candy spot {of professional} acquaintances bordering on associates. That’s the kind of one that you make some extent to seek out if you end up at a convention and get a drink with even in case you aren’t at that second writing a paper collectively. It’s that one that you shared a bit of about your non-public life with if you had been on a automotive journey collectively to the airport. It’s that one that you textual content memes of Beyonce giving out excessive fives for no good cause. It’s that individual you wish to ship a be aware to at school saying “Will you be my good friend? Circle sure or no”.
So the concept of this podcast is that she and I are going to increase an outdated research of ours with Jason Lindo and Andrea Schlosser printed within the Journal of Human Sources known as “How Far Is Too Far?” It studied what occurred when Texas handed HB2 in 2013 and almost half the state’s abortion clinics closed in a single day. We used the sudden, geographically uneven adjustments in driving distance to the closest clinic to estimate the causal impact of entry on abortion charges. The punchline was that distance issues, the consequences are non-linear, and congestion on the surviving clinics issues too.
However what we wish to do is prolong the analysis design in a few methods. First, we wish to research the impact that the abortion clinic closures had on marriage. Whereas Caitlin has studied the impact of abortion entry on marriages, nobody has take a look at the clinic closures on marriage utilizing, extra particularly, the “journey distance design” as I name it. Secondly, we’re going to be studying easy methods to estimate remedy impact parameters, in addition to what these estimands even imply, utilizing the brand new conditionally accepted (on the AER — woo hoo fellas!) steady diff-in-diff estimator by Callaway, Goodman-Bacon and Sant’Anna estimator. This estimator already has over a thousand cites and it’s solely simply now conditionally accepted — it’s not even actually actually accepted. It’s just like the AER is saying it likes you, however does it actually actually such as you? Not till it’s accepted you does the AER actually actually such as you. Proper now it’s a conditional settle for which is extra like a situationship. Anyway, I’m rooting that these two get hitched, and so we’re going to be utilizing their estimator with this journey distance design to estimate a bunch of estimands that we’re going to study collectively. In order that’s enjoyable.
After which third, and perhaps the goofiest of all — Claude Code. We’re going to do all of this utilizing Claude Code. The hope being that we are able to wrap our arms round simply easy methods to use this factor to do good, and never evil. And I feel that is the funnest (most enjoyable?) half as a result of Caitlin might be the extra pessimistic in the direction of AI, whereas I’m essentially the most optimistic, which on common means we’re aloof to AI. And Claude might be going to generally agree with me, generally with Caitlin, and generally simply wish to say all of us have an incredible level. Anyhow, we’re going to be doing this mission collectively utilizing Claude Code in order that listeners and viewers can higher see how we use Claude Code for sensible empirical analysis, and the way we go about making an attempt to get it to not leap the electrical fence, or if it does, not trigger mayhem.
However as I mentioned, Caitlin and I’ve very completely different priors on this. I’m the AI optimist and she or he’s the AI skeptic. Whereas we’ve each been utilizing Claude Code for months, and we’ve each seen what it will possibly do, and we each agree we’re within the early innings of one thing that essentially adjustments how analysis will get finished, I feel we each have elementary opinions and issues that generally overlap with one another and different occasions don’t.
However she is, I feel like me, curious to a fault. She wouldn’t be doing this if she weren’t — however she thinks AI is, in her phrases, an existential risk to humanity. And he or she will not be being dramatic. She means it. And that’s not an unusual fear amongst individuals, neither is it an unusual place to take that folks concurrently are indignant or upset about AI and wish to higher perceive Claude Code’s utility for sensible empirical analysis. That’s simply the occasions that we’re in that each of these might be true on the similar time for a similar individual. She’s the individual on the desk asking the laborious questions on what occurs when these instruments get ok that the verification drawback turns into the solely drawback.
So you’ve gotten one one that thinks that is going to be unimaginable and one one that thinks it’d finish civilization, and we’re each utilizing the identical instrument to do the identical mission. That rigidity is actual, it’s productive, and it’s a part of what you’ll hear.
And right here’s the factor about podcasting with Claude Code operating within the background: there’s lots of time whereas it’s working. It’s studying information, writing scripts, compiling issues, operating pipelines. And through that point, Caitlin and I are speaking. About AI, about science, about what we’re seeing in actual time on the display, in regards to the mission, about whether or not what simply occurred was spectacular or terrifying or each, or simply about life, in regards to the that means of being a researcher, about our worries and hopes and the place, and so forth. And we’re joking round and bantering.
It’s like The View if The View had two economists gazing a terminal.
Episodes will drop as we work by means of the mission. Some shall be knowledge work — the sort of session the place we’re elbow-deep in county FIPS codes and file format inconsistencies. Some shall be methodological — working by means of the continual diff-in-diff framework, determining what the figuring out assumptions really require. Some would be the conversations that occur in between — about AI, about the way forward for empirical analysis, about what it means to do science in public.
I don’t know what number of episodes this shall be. I don’t know what we’ll discover. I don’t know if the wedding consequence shall be a null or one thing actual or one thing we are able to’t interpret. As they are saying in remedy, it’s in regards to the journey not the vacation spot! This podcast is in regards to the journey, which is to say it’s in regards to the pleasure researchers get from doing analysis, not essentially from finishing it. And it’s a podcast of two individuals speaking whereas they do it.
The Mixtape with Scott is again. Season 5. The Odd Couple. That includes Caitlin Myers. We’re making the sausage, and also you’re invited to observe
