Michael Wiebe writes:
I have several new replications written up on my site.
Moretti (2021) studies whether larger cities drive more innovation, but I find that the event study and instrumental variable results are due to coding errors. This means that the main OLS results should not be interpreted causally.
Atwood (2022) studies the long-term economic effects of the measles vaccine. I run an event study and find that the results are explained by trends, instead of a treatment effect of the vaccine.
I [Wiebe] am also launching a Patreon, so that I can work on replications full-time.
Interesting. We’ve discussed some of Wiebe’s investigations and questions in the past; see here, here, here, and here (on the topics of promotion in China, election forecasting, historical patents, and forking paths, respectively). So, good to hear that he’s still at it!
Twitter thread on Moretti 2021 here: https://twitter.com/michael_wiebe/status/1749462957132759489
(Since Michael Wiebe’s site doesn’t allow comments.) I haven’t read though this, but at first glance at the “larger cities drive more innovation” analysis it seems important; it’s wonderful that you (Wiebe) are doing this! Aside from its technical merits, though, you might want to make the actual text of what you’ve done more inviting, with a clear non-technical abstract of what you’ve done and why the reader should care. Also, there’s a lot of terminology that’s (needlessly?) hard to follow; by paragraph 2, not knowing what “2SLS” is, I can’t continue. Especially given that you are, I think, an “outsider,” maximizing the digestibility of what you’re doing will be important for making an impact. Good luck!
Thanks for the feedback! Do you find the twitter thread more comprehensible?
https://twitter.com/michael_wiebe/status/1749462957132759489
This helps, but I still don’t think it clearly states the issue or that it draws people in. Keep in mind, of course, that I’m not your target audience and my perspective may be irrelevant. But whoever your target audience is, even if they read Moretti (2021) so frequently that it’s on the top of their minds, they have a thousand competing demands on their attention.
How about as your intro or your first tweet in the series something like: “What drives geographical hubs of innovation? Understanding this can shape policies that spur economic and technological growth. Moretti (2021) analyzed data on inventions and concluded that geographical agglomeration of inventors gives rise to a disproportionately high increase in inventions [or whatever?]. This is, however, wrong [maybe too blunt]. Reanalyzing the underlying data, I show that the event study and instrumental variable estimates that are necessary for a causal interpretation are caused by coding errors. Fixing these errors, the large inferred effect disappears. Here, in addition to I describing the data, its analysis, and the resulting conclusions, I guide the reader through code that illustrates these methods [or something like that].”
Of course, this is just my quickly written stab at how I’d approach it, and again, I’m not an economist! But perhaps it may be of use. Good luck!
2SLS is usually two-stage least squares (also stylised as TSLS). It is the linear version of general case instrumental variables from econometrics (https://en.wikipedia.org/wiki/Instrumental_variables_estimation). To be fully honest, I have never seen any other way for estimating instruments than with 2SLS. But theoretically, it should be possible to estimate instruments in other ways.
Chapter 14 of McElreath’s Statistical Rethinking has a section on fitting a Bayesian instrumental variable regression as a multivariate model. The book isn’t freely available online, but Solomon Kurz has an online “tidyverse” translation of the book, where you can see the model described mathematically and in R/brms code here: https://bookdown.org/content/4857/adventures-in-covariance.html#instrumental-variables.