@@HectorKenzo Happy yo hear you enjoy the content. I'm working on videos for undergraduate econometrics, but it will take a while until I upload them.

that just means a percentage - t%. If it was lumpsum, i guess it would just be a plain number like 500$ and not a % of the price.

@atithipatel1243 fair , thanks

Hello! The idea behind SC is quite simple: one unit got treated, and we let the computer choose units that had not been treated but had similar outcomes before the treatment kicked in. So, in the case of Spanish regions, we had one unit that was affected by terrorism (the Basque Country). Then, we let the algorithm choose regions with a similar GDP path to the Basque country before the start of the terror. The weighted average of GDP of these control regions -- in this case only two, Madrid and Catalunya -- is the synthetic Basque Country. It gives you the path of GDP that the Basque Country would have had in absence of the terror. The weights are determined by the algorithm. As a researcher you have to make a few choices: 1) what is the treated region -- that's usually determined by your research question, 2) what are potential control regions (donor pool), 3) based on what variables do I want the algorithm to construct the synthetic control. The rest is done by the algorithm.

@@ben_elsner ah i see! this explanations mean a lot to me, thankyou so much Mr. Elsner.

@@ben_elsner do potential control regions also determined by the algorithm? or we can include regions that have similiar outcome with treated before the treatment just like in DiD? is the similiar outcome enough or need another characteristic too? for example i want to evaluate a special policy in a province in Indonesia, should i include all of another province (indonesia has 34 provinces) become the donor pool? or just need some (similiar outcome, and control variabel).

@@ervinamunthe4797 no it is up to the researcher to choose which potential control units enter the donor pool. There are no set rules for how to do that. If you choose potential control units that are fundamentally different from the treated unit, the algorithm will not choose that as a control unit.

@@ervinamunthe4797 I would include all to begin with and let the algorithm choose which ones are getting included

@@shinaction2666 SUTVA should be in LATE II, i.e. in the next video

@@ben_elsner Thank you very much! Will check it out!

Thanks. Great name btw

I wasn't aware of that paper but it looks interesting. We're all scientists, so we should criticise one another's work. However, one should also assess a paper based on when it was written. A paper published in 2014 was probably written in 2010, which means that many of the new DiD methods were not there yet and the problems not fully known. Angrist & Pischke came out in 2009; so that's a good reflection of the state-of-the-art knowledge of the time. Much of the newer DiD literature started around 2018. Also, you don't know what additional results the author showed to the referees before the paper was accepted. No paper is perfect and given that this one was published in a very good journal, it probably has its merits. If you were to write such a paper in 2024, you would have to write it differently and use different techniques, but the same will be true 10/20/30 etc years from now. Methods change over time, as does the way we write papers.

Okay thank you for your quick response. What do think about the Endogeneity problem inherent in this paper. Could one make a point concerning Simultaneity bias? Or in general is the strict exogeneity assumption valid? One could very elegantly circumvent those issues with a IV approach. But I have a hard time finding such an IV that stays exogenous but is no weak instrument..

@@tobiwagner338 Apologies, I can't comment on single papers. In general, a term such as "strict exogeneity" is not helpful for (most) causal inference. That's a statistical term. The identification assumptions with DiD are parallel trends and no anticipation. If both hold (and there is no heterogeneous treatment effects or staggered rollout ;)) the DiD model identifies the ATT. Simultaneity can also be a problem (because it is unclear what the treatment is!), but solving this with IV is tricky because good IVs are hard to be found.

That's one way of using DAGs, correct. It is difficult, however, to include all possible relationships in a DAG, which is why we often use DAGs to clarify why we should or shouldn't adjust for certain variables (i.e. distinguish between confounders, mediators and colliders), but we would exploit natural experiments that cut through most the confounders. For a more detailed explanation I recommend Nick Huntington-Klein's book The Effect. He explains in great detail and with many examples how one uses DAGs in practice.

@@ben_elsner Tyvm

Hi! The J-Test (Sargan-Hansen) is not applicable in most applications because we only have one instrument for one endogenous regressor. In that case (which is called "just identified"), one cannot test whether the instrument is exogenous. All we can do is bring good arguments why this might be the case (and remember: you need to bring good arguments in favour of conditional independence AND the exclusion restriction). Tests for overidentifying restrictions like Sargan-Hansen can be used in overidentified models, i.e. in models where the number of IVs exceeds the number of endogenous regressors (as is the case in Hansen's GMM estimator). In that case, one needs one instrument that is definitely exogenous and then one can test whether the other m-1 instruments are exogenous as well. But that has very few applications in contemporary applied micro. I would not recommend using any test where the null hypothesis is that a regressor or instrument is exogenous (such as Hausmann or Sargan-Hansen). You should assume that the regressor of interest IS endogenous and then explain why in your empirical setting, under certain conditions, it can be considered exogenous. I hope this helps. Ben

@@ben_elsner Thank you for your kind reply! So, the bottom line is that the J-test assumes that one instrument MUST be exogenous, and it tests that other instruments are also exogenous as long as at least one instrument IS exogenous. Thank you! I understand!

@@Ohrobo yes exactly. You need one instrument that is valid beyond any doubt and then you can test the exogeneity of the remaining instruments. That is almost impossible in most settings (with the exception of some combined IVs like Shift-Share IVs).

how did it go

Happy to hear that! There is more to come over the next few months...