If the theory is wrong it should be retracted
Conditional on the updated Comment on arXiv, I guess the Comment would be publish and Kasy's paper would not be retracted.
The updated Comment has turned into a more independent note after they contacted Kasy. So basically Kasy asked the authors to make it more independent and give more credit to the original paper for having putting forward the idea, even if almost whole theory was wrong.
I told you that they care.
My guess is that they already submitted to ECMA or already contacted Kasy. The rule says that in these cases the authors have to attach the correspondence from the authors of the original paper. The process just takes some time.They're not people who care about a publication in ECMA
This is probably not how it will work. Now econ may really be entering in a post truth world, or returning in the soviet union area, where what matters is how important you are and not how true, correct and valid is what you say.
If the theory is wrong it should be retracted
Conditional on the updated Comment on arXiv, I guess the Comment would be publish and Kasy's paper would not be retracted.
The updated Comment has turned into a more independent note after they contacted Kasy. So basically Kasy asked the authors to make it more independent and give more credit to the original paper for having putting forward the idea, even if almost whole theory was wrong.
ECTA will never retract it. It would be an admission of a poor refereeing and editorial process. They will veil it.
If the theory is wrong it should be retracted
Conditional on the updated Comment on arXiv, I guess the Comment would be publish and Kasy's paper would not be retracted.
The updated Comment has turned into a more independent note after they contacted Kasy. So basically Kasy asked the authors to make it more independent and give more credit to the original paper for having putting forward the idea, even if almost whole theory was wrong.
ECTA will never retract it.
This is probably not how it will work. Now econ may really be entering in a post truth world, or returning in the soviet union area, where what matters is how important you are and not how true, correct and valid is what you say.
If the theory is wrong it should be retracted
Conditional on the updated Comment on arXiv, I guess the Comment would be publish and Kasy's paper would not be retracted.
The updated Comment has turned into a more independent note after they contacted Kasy. So basically Kasy asked the authors to make it more independent and give more credit to the original paper for having putting forward the idea, even if almost whole theory was wrong.
I take you haven't been around long enough to know that the way it used to work still is how it works
What I mean is that it is the first time I saw somebody behaving as Sokal in the Sokal affair, without realising it of course, and all the concerned people reacting in plain sight, very openly. This puts Econometrica in the position of Social Text. The level of hypocrisy which is in action is quite impressive, just seeing how many people are trying to put back the arXiv note in the right track. Usually the low side of the profession was less easy to catch.
This is probably not how it will work. Now econ may really be entering in a post truth world, or returning in the soviet union area, where what matters is how important you are and not how true, correct and valid is what you say.If the theory is wrong it should be retracted
Conditional on the updated Comment on arXiv, I guess the Comment would be publish and Kasy's paper would not be retracted.
The updated Comment has turned into a more independent note after they contacted Kasy. So basically Kasy asked the authors to make it more independent and give more credit to the original paper for having putting forward the idea, even if almost whole theory was wrong.I take you haven't been around long enough to know that the way it used to work still is how it works
http://www.homepages.ucl.ac.uk/~uctptk0/Research/kitagawa_tetenov_ecta2018.pdf?It seems to me that AT and TK have a recent ECMA paper on a related topic, and that there were other econometricians working in the area. It is not clear that it would have been necessary to send the paper to somebody from CS/ML, although clearly a good idea.
Kasy's proofs are short and seemingly very simple. They basically just refer to results in the CS/OR literature. The problem is that ECMA seemingly didn't ask people from those communities to review (and seemingly didn't care about novelty as long it "hadn't yet been published in econ journals").
It is indeed a quite common topic among Urobros. This one should appear in JASA according to the website of one of the authors.
https://arxiv.org/abs/1812.09408#
It seems to me that AT and TK have a recent ECMA paper on a related topic, and that there were other econometricians working in the area. It is not clear that it would have been necessary to send the paper to somebody from CS/ML, although clearly a good idea.Kasy's proofs are short and seemingly very simple. They basically just refer to results in the CS/OR literature. The problem is that ECMA seemingly didn't ask people from those communities to review (and seemingly didn't care about novelty as long it "hadn't yet been published in econ journals").
http://www.homepages.ucl.ac.uk/~uctptk0/Research/kitagawa_tetenov_ecta2018.pdf?
So what you say is that a certain number of incorrect theorems must be accepted by the community? I would agree to this. Everyone who is interested and has the skills can write a correction of such results.The truth is, unfortunately, the standard of theory reviewing remained the same or worse from the 1980s when the complexity of math and the length/number of proofs was a fraction of todays. People complain that reviewing takes too long - which it does - but it's a pretty tricky process. I had a paper published at a top 5 journal in which I caught an incorrect proof for an inconsequential proposition before the final reviewing round. It would've gone through with the mistake in it. None of the almost 10 referees said anything. There's an increasingly smaller number of qualified theorists in the field and it may well take months for a decent theorist to verify proofs in a paper.
It's pretty tricky. Either you create a costly pre-committement device or wrong proofs are really the same as coding errors and the cost needs to fall on the reviewing process. Once I got a paper to referee that had 83 pages of proofs. The order of the proofs was completely wrong (e.g., lemma that wasn't lemma, result in proposition 10 being used in prop 2...). I told the editor to send to someone else or send it back to the authors to re-arrange the whole thing. It would take me hundreds of hours to figure out what was going on there. Maybe theory journals should have proof standards like applied outlets are implementing with code? Most theory papers have one or two main theorems that, in case of being mistaken, the whole paper is wrong. Clearly say all results that are used in the theorem and someone at the journal needs to go through that to see if all pieces are there.
Comments on the quality of referees remind me of this:
https://faculty.wcas.northwestern.edu/~bhs675/PPIC.pdf
I wonder if the original referees even thought they understood the proofs.
So what you say is that a certain number of incorrect theorems must be accepted by the community? I would agree to this. Everyone who is interested and has the skills can write a correction of such results.The truth is, unfortunately, the standard of theory reviewing remained the same or worse from the 1980s when the complexity of math and the length/number of proofs was a fraction of todays. People complain that reviewing takes too long - which it does - but it's a pretty tricky process. I had a paper published at a top 5 journal in which I caught an incorrect proof for an inconsequential proposition before the final reviewing round. It would've gone through with the mistake in it. None of the almost 10 referees said anything. There's an increasingly smaller number of qualified theorists in the field and it may well take months for a decent theorist to verify proofs in a paper.
It's pretty tricky. Either you create a costly pre-committement device or wrong proofs are really the same as coding errors and the cost needs to fall on the reviewing process. Once I got a paper to referee that had 83 pages of proofs. The order of the proofs was completely wrong (e.g., lemma that wasn't lemma, result in proposition 10 being used in prop 2...). I told the editor to send to someone else or send it back to the authors to re-arrange the whole thing. It would take me hundreds of hours to figure out what was going on there. Maybe theory journals should have proof standards like applied outlets are implementing with code? Most theory papers have one or two main theorems that, in case of being mistaken, the whole paper is wrong. Clearly say all results that are used in the theorem and someone at the journal needs to go through that to see if all pieces are there.
Comments on the quality of referees remind me of this:
https://faculty.wcas.northwestern.edu/~bhs675/PPIC.pdf
I wonder if the original referees even thought they understood the proofs.
And this: https://www.sciencedirect.com/science/article/abs/pii/S0304407619300181
So what you say is that a certain number of incorrect theorems must be accepted by the community? I would agree to this. Everyone who is interested and has the skills can write a correction of such results.The truth is, unfortunately, the standard of theory reviewing remained the same or worse from the 1980s when the complexity of math and the length/number of proofs was a fraction of todays. People complain that reviewing takes too long - which it does - but it's a pretty tricky process. I had a paper published at a top 5 journal in which I caught an incorrect proof for an inconsequential proposition before the final reviewing round. It would've gone through with the mistake in it. None of the almost 10 referees said anything. There's an increasingly smaller number of qualified theorists in the field and it may well take months for a decent theorist to verify proofs in a paper.
It's pretty tricky. Either you create a costly pre-committement device or wrong proofs are really the same as coding errors and the cost needs to fall on the reviewing process. Once I got a paper to referee that had 83 pages of proofs. The order of the proofs was completely wrong (e.g., lemma that wasn't lemma, result in proposition 10 being used in prop 2...). I told the editor to send to someone else or send it back to the authors to re-arrange the whole thing. It would take me hundreds of hours to figure out what was going on there. Maybe theory journals should have proof standards like applied outlets are implementing with code? Most theory papers have one or two main theorems that, in case of being mistaken, the whole paper is wrong. Clearly say all results that are used in the theorem and someone at the journal needs to go through that to see if all pieces are there.
Comments on the quality of referees remind me of this:
https://faculty.wcas.northwestern.edu/~bhs675/PPIC.pdf
I wonder if the original referees even thought they understood the proofs.
most theory papers are much longer than they should be. I would never say the proof is difficult to read. But it is certainly too tedious and time consuming for referees. One more point on these sort of theory papers: unlike econometric theory, no one outside the theory community actually applies their theorems. So they are mostly harmless because they are mostly useless.
I think it is too pessimistic. Look for instance on how the notion of correlated equilibrium is attracting some interests from econometricians and CS researchers. The fact that the bandit literature, which can be very theoretical, is attracting attention is also a good sign. There are however two important issues: 1/ it takes time to understand this kind of literature, and 2/ you need a good formal training to be able to, and it is not given to many people, due to the quantitative importance of other fields which may be less demanding from that respect.
most theory papers are much longer than they should be. I would never say the proof is difficult to read. But it is certainly too tedious and time consuming for referees. One more point on these sort of theory papers: unlike econometric theory, no one outside the theory community actually applies their theorems. So they are mostly harmless because they are mostly useless.
^ the proofs are tedious because the authors simply are not aware of the theorems/propositions that have been proven in other fields e/g/. maths/stats. Other times, to show that their study make contribution the authors just (intentionally) reinvented the wheel!
So what you say is that a certain number of incorrect theorems must be accepted by the community? I would agree to this. Everyone who is interested and has the skills can write a correction of such results.The truth is, unfortunately, the standard of theory reviewing remained the same or worse from the 1980s when the complexity of math and the length/number of proofs was a fraction of todays. People complain that reviewing takes too long - which it does - but it's a pretty tricky process. I had a paper published at a top 5 journal in which I caught an incorrect proof for an inconsequential proposition before the final reviewing round. It would've gone through with the mistake in it. None of the almost 10 referees said anything. There's an increasingly smaller number of qualified theorists in the field and it may well take months for a decent theorist to verify proofs in a paper.
It's pretty tricky. Either you create a costly pre-committement device or wrong proofs are really the same as coding errors and the cost needs to fall on the reviewing process. Once I got a paper to referee that had 83 pages of proofs. The order of the proofs was completely wrong (e.g., lemma that wasn't lemma, result in proposition 10 being used in prop 2...). I told the editor to send to someone else or send it back to the authors to re-arrange the whole thing. It would take me hundreds of hours to figure out what was going on there. Maybe theory journals should have proof standards like applied outlets are implementing with code? Most theory papers have one or two main theorems that, in case of being mistaken, the whole paper is wrong. Clearly say all results that are used in the theorem and someone at the journal needs to go through that to see if all pieces are there.
Comments on the quality of referees remind me of this:
https://faculty.wcas.northwestern.edu/~bhs675/PPIC.pdf
I wonder if the original referees even thought they understood the proofs.
What font is this?