Last updated: November 2022
This repository tracks the recent developments and innovations in the Difference-in-Difference (DiD) literature. It serves two purposes. First, it is an organized collection of various bookmarks from Twitter, GitHub, YouTube etc. Second, it aims to present the different packages from an end-user’s perspective. This part has to do with how to apply these methods in day-to-day applied research. On the theory side, several really useful resources are listed in the Resources including workshops and notes by some of the key authors leading the development in this field. Please refer to this section if you want a deeper theoreical understanding.
The contents of this page will get intermitent updates as necessary. It might contain errors, or links might get broken, or packages get updated, or new papers are not listed. Therefore, please give feedback by opening a Pull Request (PR) or opening an Issue. The aim of this repository is to collectively build notes and a code base that we can all use.
Some thoughts below from my own perspective (these are also subject to evolve over time):
Several DiD innovations came out simultaneously in 2020 and 2021 with some staggered roll-outs in 2022. At the heart of this new DiD literature is the premise that the classic Two-way Fixed Effects (TWFE) model can give wrong estimates. This is especially true if the treatments are heterogeneous (differential treatment timings, different treatment sizes, different treatment statuses over time) that can result in “negative weights” contaminating the ATE.
Recent innovations like the Bacon decomposition help us unpack the weights of the different combinations of different treated and untreated cohorts. The new DiD methods introduce various DiD estimation techniques that “correct” for TWFE biases. Within these packages are different ways of handling parallel trends, negative weights, covariates, controls, etc. This is done using different methods ranging from bootstrapping, inverse probability weights, matching, influence functions, imputations, etc. The packages are constantly being improved and implemented in different languages. Currently, it is also not very clear which methods/packages work best for which problems. Hopefully, more will be written on comparing the utility of each estimation technique by those who know this stuff better. At the time of writing this, new papers and packages are still being released. Hopefully, we will also start seeing more applications and replications that can help us understand the nuances across the different methods.
In 2022, several review papers came out, that summarize the state-of-the-field really well. They are a good starting point to familiarize oneself with the methods and are marked in the literature section.
The aim of the code repository is to help readers navigate the code structure, and syntax usage for different packages. Therefore, this repository is periodically updated. If you want to report errors and/or contribute, then please open an issue, or start a discussion, or e-mail me at firstname.lastname@example.org.
If you use this repository and find it helpful, acknowledgements and/or citations will be highly appreciated.