Upcoming Workshops on Causal Inference, DiD and Synthetic Control

Announcement

Some time late 2019, I began conducting workshops on causal inference to departments. I would be flown out and present for 2, sometimes 3, days on the material that is in my book, Causal Inference: the Mixtape. I guess I’ve done a couple dozen of these, both on-site, and online since then, including a large one on difference-in-differences this summer called CodeChella. I’ve gotten good feedback and am continuing to be invited to do these for departments as well as firms. I’ve done these at places like Rotterdam, Catholic University of Uruguay, University of Wyoming, Facebook, University of College London, University of Pennsylvania and dozens more. They are a wonderful time for me, I love doing them, and I’ve heard good things about them from faculty and students. And if you are interested in having me come out to do one for you, be it an academic department (any department) or any organization, feel free to contact my assistant, Emmi Scott at emmiscott15@gmail.com, and she can set it up.

But not everyone can have me come out. And with the stability achieved in video conference technology like Zoom, and the norms around doing online seminars having been firmly established by COVID, many workshops have moved online. Statistical Horizons, for instance, does a lot of these online. So it’s not so uncommon to see continuing education courses in statistics and programming conducted online. Which brings me to my point.

Starting in 2022, I’m going to be hosting online workshops in causal inference, as well as new developments in difference in differences and synthetic control. These will follow a price discrimination model wherein students, predocs and postdocs pay discounted prices whereas non-students pay higher prices. If someone is unable to pay these prices, there are scholarships as well as a sliding scale fee that you can work out with my assistant Emmi. Part of my philosophy is and always be that it is most important that causal inference methods, both old and new, be distributed on a regular basis to anyone and everyone who needs it. So if you need this material, we can work it out. Below is a description of the workshops, but if you are interested, please fill out the information at this google form so that I can contact you and send you the syllabi as well as start communications to learn more about your preferences and whether we can work something out.

The first workshop will likely be in mid-January on causal inference and will go two consecutive weeks on a Friday and Saturday. Throughout the year I will also provide them on international times assuming there is demand for that. Feel free to share this to anyone you know who may be interested in learning this material.

Workshop 1: Causal Inference and Research Design

Causal inference is a specialization within economics and statistics that grew out of the labor economics tradition to evaluate the causal effects of programs. While physical randomization was widely known to yield unbiased estimates of causal effects, it was not often used in economics.  A wave of new labor economists starting in the late 1970s and mid 1980s changed that as they pushed for focus on exploiting quasi random assignment in natural experiments or through imposed modeling assumptions on counterfactuals.  This work in conjunction with pioneering work in econometrics led to the sharpening of such research designs as instrumental variables and difference in differences.  This workshop will cover foundational elements of modern practices of causal inference such as the potential outcomes model as well as discuss in detail the most common designs: regression discontinuity, instrumental variables, difference in differences, comparative case studies using synthetic control and if time permitting matching.  It will be accompanied by efforts to introduce students to basic practices in programming as well as good research practices more generally.

Aims of the class:

  1. To help students become more familiar with the field of causal inference

  2. To empower students to apply research designs more competently to their own research

  3. To direct students towards better programming practices so that they are better able to perform quantitative forms of research

Daily Structure

This is a 4-day workshop.  The goal of the workshop is for students to gain enough knowledge from the lectures and experience from the programming activities that they become confident and capable enough to implement and interpret these methods in their own work, as well as continue to learn this new material on their own after the workshop concludes.  Each day lasts 8 hours with 4 hours of lecturing, 2 separate 75 minute “coding together” sessions, and the remainder are breaks and lunch.  We will be strict about holding the 2.5 hours of coding and therefore may sacrifice on breaks if we are running behind.

Topics:

We will cover basic causality concepts on counterfactuals using potential outcomes and causal graphs (DAGs), randomization inference, regression discontinuity, difference in differences, synthetic control, instrumental variables and matching.

Workshop 2: New Developments in Difference in differences and synthetic control

When researchers are not be able to field randomized experiments to study the causal effects of large social programs due to their size, associated costs, feasibility and ethical constraints, they often rely on natural experiments such as law changes or natural disasters.  The most popular research designs for estimating the causal effects using such natural experiments are the difference-in-differences design and synthetic control estimation. Both difference-in-differences and synthetic control have evolved considerably over the last several years, both in terms of econometric theory and software implementation. This workshop will review this emerging work covering both the intuition behind the statistical models and the technical details of the models themselves using lectures, discussion and group exercises using R and/or Stata.

Daily Structure

This is a 3-day workshop.  The goal of the workshop is for students to gain enough knowledge from the lectures and experience from the programming activities that they become confident and capable enough to implement and interpret these methods in their own work, as well as continue to learn this new material on their own after the workshop concludes.  Each day lasts 8 hours with 4 hours of lecturing, 2 separate 75 minute “coding together” sessions, and the remainder are breaks and lunch.  We will be strict about holding the 2.5 hours of coding and therefore may sacrifice on breaks if we are running behind.

Aims of the class:

  1. To help students become more familiar with the field of causal inference

  2. To empower students to apply research designs more competently to their own research

  3. To direct students towards better programming practices so that they are better able to perform quantitative forms of research

Topics:

We will cover basics of difference-in-differences, the inclusion of covariates, issues and biases with differential timing, alternatives to two-way fixed effects, synthetic control, matrix completion with nuclear norm regularization, synthetic difference in differences, and augmented synthetic control.