Apologies for going AWOL for a week. I think that I hit a bit of a wall after my trip to Germany is all I can deduce. When I got back, and I couldn’t get the interview with Richard Blundell ready in time to post, I had to put it off a week and that seemed to kick me out of a groove I was in. I also started a four day workshop spread over two weekends and was trying to get ready for that. One thing led to another and I seemed to want to rest and focus on my classes and getting caught up. I suspect I’m setting too many meetings on too many days on too many irregular hours, also. But I’m pushing through and coming up for air.
So what is this post about? I wanted to use todays substack to outline some thoughts I’m having about Mixtape Sessions, my online causal inference education platform. Don’t hold me to any of this please as I just wanted to articulate it, not necessarily commit to anything. But I’m just wanting to think out loud while I am so update readers and supporters on how Mixtape Sessions is doing, where it’s come from, and (maybe) where it is going. A lot of what I am going to write here will blend my personal life and my professional life, but the truth it, they’re at this point fairly inseparable. I have a sense of mission about educating and helping people learn and apply causal inference methodologies, and it sometimes makes it not very clear where my personal and professional life start and end. But let me try to focus and finish this as I’ve been writing it now for days and I doubt it’s worth the time I’ve spent on it.
What is Mixtape Sessions?
Mixtape Sessions is a platform where I and others teach people about causal inference and econometrics. I started it in November 2021. I reached out to Kyle Butts (University of Arkansas’s newest assistant professor!) and asked him if he could help me create the website and as always, Kyle dove in and made something exciting that exceeded my imaginations. That something was a website (here), a GitHub repository (here) and a branded aesthetic (including an icon) for all the slides. It’s a platform where I, but also around a dozen other professors, have met with guests and taught workshops on a range of topics around econometrics from the core causal inference methodologies in the design tradition like difference-in-differences and instrumental variables as well as recently structural demand estimation. The dream was that it could function as a hub connecting speakers and presenters with eager students and scientists of all types.
In addition to what I said, it’s also a “branded” collection of workshops, with a particular visual style that Kyle Butts created about causal inference and applied econometrics topics taught by myself, as well as top econometricians and experts from around the world. The orange Sun icon with the time series graphic is something Kyle created, as is the website, the organization of the GitHub repos, and the shirts he made for us. I can’t do what Kyle can do. He’s brilliant. And while I think we both come at this “give away knowledge” model from a different background, we still seem to see the platform and the goals more or less the same. So when I start trying to use price discrimination and low prices to find every conceivable human alive to match them with a class or a speaker, Kyle never bats an eye.
My motives for doing it all are complex. I barely understand them and they seem to be a combination of three or four different things. But I do think it is true that deep down, at my core, I long to help someone learn this so that they can get to their goals as a researcher faster than it took me, without nearly as many trips and falls. And I also think with everything inside me that I can help people love it. Not just learn it but love it. Love causal inference. Love potential outcomes. Love econometrics. Love numbers. And I ask only that they indulge my desire to play rap music at the intermissions and watch me cue up video clips with economists I’ve interviewed as I try to articulate a story of the paradigm shifts that the laureates and others wrought.
But the actual effort, Mixtape Sessions, started in November 2021. The first workshop was January 2022, but I posted on Twitter plans to do it in the early winter of 2021 in response to the demand I sensed for the new difference-in-differences material that had been booming, as well as in response to the popularity of my book, Causal Inference: the Mixtape, which had been published by Yale Press earlier that year in January 2021. Covid had caused the world to shut down but academia continued connecting to one another through social media and Zoom. I’d already had success that summer using Twitch to teach workshops on diff-in-diff, and so I decided that winter to do something recurring and permanent. And that was the idea for Mixtape Sessions.
The mission of Mixtape Sessions is intended to be a hub between those who want to learn this material and those who know it. It seems like these hubs have become important — maybe moreso than was once the case. I was told recently by a senior health economist that never before had it been true that the time from when a new method appears to when it is required has shrunk this short. We saw this I think with difference-in-differences in particular — the growing saliency around the fact that two-way fixed effects was biased with heterogenous treatment effects was really a locomotive that smashed through economics from 2018 to 2021. It was likely sped up via Twitter where many econometricians posted regularly about the biases, did lengthy explainers (some myself), and engaged in a novel kind of mass communication that I don’t think we’d seen before with econometricians. These were largely younger econometricians, many recent PhD graduates and assistant professors, and #EconTwitter was at the time a vibrant community of sorts, well connected, largely flat and egalitarian, and filled with lurkers as well as vocal participants.
But I still felt that there was more that needed to be done that was more traditional. Twitter threads were not enough. And I had hoped to have it done so that I could include it in the book, but the timing of everything made it impossible as I was already way behind schedule when I submitted the manuscript in the summer of 2020. The fact that I sought to put it in the book at all probably is an indicator of how much collective anxiety I sensed in the profession around this new material. The idea that twoway fixed effects, arguably the most common method employed, in difference-in-differences, arguably the most common design employed, under differential timing, arguably the most common type of diff-in-diff design employed, had caused in my mind a real collective emotion ranging from curiosity to outright paranoia. But everyone seemed engaged on some level back then, so while I did include a detailed discussion of Andrew Goodman-Bacon’s now classic on difference-in-differences, I had not really done much to flesh out that new material other than a fairly short cursory glance at Callaway and Sant’Anna and matrix completion with nuclear norm regularization, and even then it was inadequate.
So after I submitted the manuscript in late summer 2020, I switched to reading the diff-in-diff papers more closely. Without the stress of a missed deadline looming over me, I felt freed up to study this material. But the problem was there was a decent amount of it, it was technical, and I wanted a map to navigate it. I wanted a map for myself, but I also wanted a map for others. They were fairly correlated desires given how popular difference-in-differences was and how popular the Mixtape was as well. It felt like my desire to help bridge the gaps in understanding causal inference was opening up for me with difference-in-differences and as I had just barely missed the chance to include it in the Mixtape, the material was needing demystification.
So early 2021 became a period of intense learning and reflection. We were still in the thick of the pandemic. Online learning was becoming a major part of everybody’s life, and I was seeing more and more online workshops, too. I had begun teaching my own causal inference workshops for free to students semi-regularly at this time. But I was really wanting to do more to help people learn the new difference-in-differences papers. So I began writing substacks and Twitter threads on difference-in-differences with complete abandon. I also publicly offered a bounty for the first person to create a working version of the Callaway and Sant’Anna estimator in Stata which ultimately led to the creation of csdid by Fernando Rios-Avila, Brantly Callaway and Pedro Sant’Anna. Sometimes if you use it, an Easter Egg will pop up and show you this:
I was compelled by a personality that got utility from helping people and got utility from scratching my own curiosity itches. Given I had developed a belief that causal inference had been transmitted from one person to another through informal networks and apprenticing rather than econometrics textbooks, I felt simultaneously that everything I learned was not going to get to the right people unless I created a bridge to them. That became very apparent when I floated the idea publicly of a workshop devoted just to difference-in-differences. The massive interest in a proposed public workshop, which later became known as CodeChella, confirmed my suspicions: the demand for understanding was not just substantial; it was urgent. Nearly 3,500 people expressed interest via a Google form I had posted (see below) which reinforced two things: 1) I was right that there was demand for difference-in-differences education and 2) if I wanted to do it I could. I still can’t believe 3500 people expressed interest.
So I posted a substack advertising it and watched as people bought tee shirts for the free lengthy workshop. And when it happened, CodeChella seemed to cause a personal revelation. First, 1,500 people attended those Twitch lectures. I filmed them and posted them to YouTube where they continue to get dozens of views daily. But the second, the event validated my belief that my style of teaching could be useful right now. So if I felt there needed to be a bridge, there could be if I wanted to be it. Second, I thought maybe demand was growing.
So, like I said, a bridge. A hub. I wanted to make a bridge to econometrics that was both broad enough to reach many and sturdy enough to bear the weight of complex concepts. And that’s where Mixtape Sessions emerged. It emerged from a belief that causal inference needed more efforts to reach people as I felt many were being left behind.
Mixtape Sessions Since 2022
Since launching Mixtape Sessions in January 2022, there have been 45 workshops, 12 different speakers, and 5,764 eager guests. The workshops have followed core topics in the causal inference toolbox like difference-in-differences, synthetic control and instrumental variables (among others) and are led not just by me, but primarily a roster of all stars including Alberto Abadie, Peter Hull, Jon Roth, Rocío Titunik, Brantly Callaway, Brigham Frandsen, Christopher Walters, Ariel Pakes, Jeff Gortmaker, Mark Anderson and Dan Rees. Attendance over the last two and a half years grew as we invited more people to teach.
Mixtape Sessions has evolved into a vibrant platform for learning and sharing about econometrics and causal inference. Our inaugural workshop on January 14th, 2022 marked my first officially branded effort into democratizing econometrics education and drew 129 attendees. It was a workshop on causal inference matching my book for the most part. But the second one in February was a four day workshop exclusively on difference-in-differences and attracted 147 participants. This past weekend I concluded my tenth difference-in-differences workshop on Mixtape Sessions and it had 242 visitors. As the pictures above show, the workshops keep getting bigger.
Indulge me for a minute, but in writing this, I dug a little more into the data on attendance. First more than 40% of the people who come to Mixtape Sessions come to more than 1 workshop. Theres some really enthusiastic supporters. One person has attended 26 workshops! Wow. 🤩
But there’s also signs of interesting bridging happening which is what I was hoping for. The biggest divide I can think of among applied microeconomists is between reduced form and structural econometrics. They are as far from one another as the East is from the west it can sometimes seem. So it was exciting to have in the workshops that have taken place a structural demand estimation workshop led by Jeff Gortmaker and Ariel Pakes a month ago. This was our most popular workshop so far with 313 people who registered. People who sign up for a workshop to learn about a core empirical model in structural IO would seem to be a pretty long throw away from the type of reduced form econometric material that are normally on here. But get this: of the 313 people who registered for Jeff and Ariel’s workshop, 116 of them had already been to a workshop on Mixtape Sessions before. Almost 40% of the people who attended a structural econometrics IO workshop were coming to it via the platform’s traditional workshops on reduced form causal inference methodologies. The hub was working — people were walking from one workshop to another, likely helped by the hopefully affordable prices.
No doubt there is more of this kind of cross-pollination that is possible but the fact that I have so much repeated interaction with the workshops, as well as people moving between topics that perhaps may not have been possible before, is a fact that I have been mulling over for a while. Some of the mixing has been between academic and non-academic attendees. I can’t show how many academic and non-academic attendees I have, as I only have email addresses, but I can at least show a little bit. My overall sense from interacting with people in the workshops has been that the platform is used more and more by non-academics. I don’t ask for a lot of data, but here’s a distribution of registered guests by their email Domain. Academic emails are 40% of guests.
From within academia, the best I can get is a count of the .edu domain names. Note these are frequencies spread over the entire 2.5 years unadjusted for which workshops they attended.
Mixtape Sessions Going Forward
My journey with the Mixtape book and the Mixtape Sessions platform were both unexpected delights. It has all been a pleasant surprise to both learn that I was right that there is demand for accessible, high-quality econometrics education and that I might be able to get involved it in so many different ways. It feels good to feel like you can give back to the community. I love economics as a science and I love the grand story of the economics community. I know many who attend aren’t economists, but I still see all this as an extension of economics given my connection to the Laureates paradigm and how it has so radically altered my worldview.
So that brings me to the future. I am contemplating a way to keep this going but in such a way that I can reallocate my time. I’ve not been able to finish the new book and other professional goals because I am doing so many workshops intensively. I never really turn any request down. And I don’t want to. This has meant a lot to me. My heart is in this. But I do feel that the schedule is a tough one. For one, my workshops are 4 days long, spread over two weekends, and they happen 8 times a year not counting travel. And that has meant not being as fully engaged in my church as I want. This church has told me they love me and want me to be part of their community. As I’ve said, they call me “Brother Cunningham” from the pulpit or during prayer and they hug me when they see me. I want to be there on Sundays. There’s only 18 people who attend, and 16 are elderly women. The church is aging. I won’t always have them. And I’m not getting younger either. So that is in my mind.
But the demand for this material I think remains high. I have a lot of things figured out and they’re working well. The roads into Europe and Asia have been in particular very rewarding. I want to help. The existence of the platform has meant someone in India can pay a dollar and come to a workshop by Alberto Abadie, Peter Hull or Rocío Titunik. A student or someone unemployed can pay $50. Someone not on a tenure track or visiting or with a high teaching load can pay $95. It isn’t an option that this not exist.
But I also want to keep learning too. I enjoyed all that time and effort I put into the book and then CodeChella and the new diff in diff material. But it’s getting time to retool again. Deep down, I love learning and putting things into my own mother tongue more than anything. I love thoughts. Thinking. Trying to push through my ignorance. It means a lot. But I can’t easily see where I’d put a fourth workshop. I’m already doing three. And yet I want to develop new workshops. I’ve found my mind thinks now in those terms too — workshops.
So in the spirit of innovation and driven by my dual passion for learning and for teaching, I'm considering a subscription-based model. This would not simply be a repository of workshops but an expansive, ever-evolving library of econometric wisdom. Imagine having at your fingertips a wealth of explainer content, code syntax tutorials, practical applications, and exercises—all designed to demystify econometrics in the unique, approachable style that Mixtape Sessions is known for.
If I made that move, it would maybe be slow. But it’s something I’m considering because maybe what would be better is something like a channel, as opposed to repeating workshops. People subscribe, they get access to whatever I have built up, until they cancel. And that’s it. I just developed new material and post it there. Tied to new papers or even old papers and just try to make it practical and accessible.
So that’s where things stand. I’m nearing the 100th episode of the podcast. This summer I go to Europe. So maybe every now and then I’ll share a little about where my head is. But for now that’s where things stand. Thanks for listening.
After attending just 1 mixtape session, I can definately say a subscription model would be invaluable !!
I haven't attended a session, but I would be highly likely to subscribe if a subscription version were available, either as student now or prof later when/if it becomes available.