Event Studies (The Effect, Videos on Causality, Ep 48)

Please visit www.theeffectb... to read The Effect online for free, or find links to purchase a physical copy or ebook.
The Effect is a book about research design and causal inference. How can we use data to learn about the world? How can we answer questions about whether X causes Y even if we can't run a randomized experiment? The book covers these things and plenty more. These videos are meant to accompany the book, although they can also be viewed on their own.
This video relates to material found in Chapter 17 of the book.
A version of this video without background music can be found here: • Event Studies (The Eff...
What if we want to get the causal effect of something that... happened? Let's start simple. Just compare what was before the effect to what came after. With some adjustments that might just be enough! Welcome to the event study.

[@wangguan1548]

Sorry, another question, does event study to be tracked on the same unit? Like in a panel survey. Can repeated cross sectional data be used for event study? thanks

[@arunbhat1784]

Good one

[@wangguan1548]

Hey, professor, do we need to have any comparison group (to treated) for event study? ( like in a DID study)

[@NickHuntingtonKlein]

@@wangguan1548 as I use the term, event studies are for cases where you are not using a comparison group. So you don't need a comparison group for this, although conversely this approach is only appropriate when you don't need one.

[@wangguan1548]

@@NickHuntingtonKlein Sorry, I may misunderstand, is it (event study) similar to interrupted time series, if without the comparison groups in this context?

[@NickHuntingtonKlein]

@@wangguan1548 correct. I recommend checking out the event study chapter in the book. www.theeffectbook.net/ch-EventStudies.html

[@wangguan1548]

@@NickHuntingtonKlein Thanks, very clear

[@tayoukachukwu3105]

Hi Nick, thanks for the video. Please I have a question. How is this event study technique different from a paired sample t-test? For example you have a sample at time t=0 and you measure the outcome mean m0. Then you administer treatment to the sample and at time t=1 you measure the outcome mean m1. Then you do a t-test on the outcome means m0 and m1 and if you reject the null you conclude that the difference in means is statistically significantly different from 0 and hence the treatment had an effect on the sample. I have always wondered whether this treatment effect can be considered causal given that we do not have a control group that represents the counterfactual. I have seen psychology researchers argue that this treatment effect may be considered causal.
From event study discussion above we have something similar in that the prediction beyond event occurrence is like the outcome mean before treatment (m0) and then we compare this to the outcome mean after event/treatment (m1). How can we call the treatment effect causal in this event study case, given that we do not have a control group or a counterfactual, just like in the paired sample t-test case? Can we really consider the prediction of the status quo before treatment as the valid counterfactual?
I hope that I am not asking for too much. I actually thought the the DID design helps to overcome this deficiency of the event study design by explicitly including a control group to act as the counterfactual and in this way better control for contemporaneous events that may contaminate the treatment. Thanks and later

[@NickHuntingtonKlein]

You could consider a paired t test of the kind you're describing as a very simple kind of event study. But usually event studies consider the time-dependent nature of the data, things like trends and careful predictions of the counterfactual more carefully. See the other videos I have on event studies, or read the chapter, for more detail

[@tayoukachukwu3105]

@@NickHuntingtonKlein Thanks for the prompt response. Much appreciated. I have read the chapter in the book and had doubts about the careful predictions used as counterfactual arguments. I think that there is a reason event studies are not as common as expected, even given their apparent simplicity, and are not usually discussed as one of the more robust causal econometrics techniques. I was actually excited seeing it as a chapter in your book, but still not convinced that it is as robust as DID as regards identifying causal effects. I may be wrong though and would like to get your view on its robustness in identifying causal effects and the case for its greater adoption. Thanks

[@NickHuntingtonKlein]

@@tayoukachukwu3105 I generally agree that it's not as robust as other methods. Whether it makes sense basically comes down to whether you think you can best estimate the counterfactual using prior trends of the treated group of interest, or using the actual trend of a control group.

[@ginaelconstelmpoubou1986]

Great