Regardless of whether Traina is right or not (and he is), my main problem with this whole thing is the power differential. JE and JdL come across as extremely petty with this behavior.
De Loecker and Eckhout go on all out attack
-
Lol, these guys are unhinged. Overselling a flawed paper, and then going ad hominem. What Traina pointed out wasn't even the only flaw in their paper.
Yeah, probably the most obvious and uncontroversial flaw is something they don't even really address in their response - the Edmond, Midrigan, and Xu point that they should be weighting by cost shares rather than sales shares. This makes a HUGE difference: the sales-weighted markup average is way higher than the cost-weighted markup average, and has seen way more of an increase.
If we were using a simple, sales/total cost definition of the markup, then it would be obvious that we should weight firm-level ratios by shares of total cost, because that would recover the right aggregate ratio. (In general, if you are averaging ratios, you weight by the denominator rather than the numerator - not hard stuff!)
It is a bit more complicated here because we are interested in markups over marginal cost, but still, Edmond, Midrigan, and Xu point out that under reasonably general conditions it is theoretically most accurate to weight by cost. Weighting by sales gives hugely inflated values.
I think the Traina critique is broadly right too, but it is a bit trickier because it gets into this accounting morass that neither JDL and JE nor most macroeconomists know anything about -- what exactly is COGS vs. SG&A, and how does it map onto our notions of variable cost? JDL and JE are certainly wrong to claim that COGS = variable cost, and their results are vulnerable to economically irrelevant accounting changes in the extent to which variable vs. fixed costs have been counted as COGS vs. SG&A over time.
This is an interesting case: I'm not sure how many times I've seen a blockbuster paper get such incisive and damning critiques before it's published. I think it would be a bit depressing, actually, if JDL and JE's paper does publish well: the message it will send is "make some bad methodological decisions, get huge sexy numbers, then oversell those huge numbers, and you will get a QJE". I think their horrifying behavior wrt Traina is just one manifestation of the ugly game that is going on here.
At some point the QJE really has to ask whether they should let this one go through.
-
I think the Traina critique is broadly right too, but it is a bit trickier because it gets into this accounting morass that neither JDL and JE nor most macroeconomists know anything about -- what exactly is COGS vs. SG&A, and how does it map onto our notions of variable cost? JDL and JE are certainly wrong to claim that COGS = variable cost, and their results are vulnerable to economically irrelevant accounting changes in the extent to which variable vs. fixed costs have been counted as COGS vs. SG&A over time.
I'm an accountant. It's doubly complicated because we don't necessarily know the macro notion of variable cost. If it's just defined as any cost that varies with output, then both COGS and SG&A have variable and fixed parts.
But if you want to know the difference between COGS and SG&A, COGS (Cost of Goods Sold) is based on the cost of the company's inventories, also known as "product" costs. From a financial accounting perspective, COGS includes manufacturing overhead, which has both variable and fixed components.
SG&A (Selling, General & Administrative) are non-product costs. It includes non-manufacturing overhead, which again has both variable and fixed components.
-
I think the Traina critique is broadly right too, but it is a bit trickier because it gets into this accounting morass that neither JDL and JE nor most macroeconomists know anything about -- what exactly is COGS vs. SG&A, and how does it map onto our notions of variable cost? JDL and JE are certainly wrong to claim that COGS = variable cost, and their results are vulnerable to economically irrelevant accounting changes in the extent to which variable vs. fixed costs have been counted as COGS vs. SG&A over time.
This is an interesting case: I'm not sure how many times I've seen a blockbuster paper get such incisive and damning critiques before it's published. I think it would be a bit depressing, actually, if JDL and JE's paper does publish well: the message it will send is "make some bad methodological decisions, get huge sexy numbers, then oversell those huge numbers, and you will get a QJE". I think their horrifying behavior wrt Traina is just one manifestation of the ugly game that is going on here.
At some point the QJE really has to ask whether they should let this one go through.It is clearly incorrect to simply treat SGA as fixed costs.
This is from Karabarbounis and Neiman:
" SG&A is described in Compustat’s data
definitions as including “all commercial expenses of operation (such as, expenses not directly related to product production) incurred in the regular course of business pertaining to the securing of operating income...” Such expenses explicitly include categories like marketing or R&D,
where it is unclear if they should be variable costs in the sense desired for markup estimation, but also includes bad debt expenses, commissions, delivery expenses, lease rentals, retailer rent expenses, as well as other items that more clearly should be included as variable costs. Most importantly, Compustat itself explicitly corroborates the blurred line between COGS and SG&A when it states that items will only be included in COGS if the reporting company does not themselves allocate them to SG&A. Similarly, Compustat does not include items in SG&A if the reporting company already allocates them to COGS." -
Everyone siding blindly with Traina clearly is not active in IO/competition policy. At basically every major conference these issues have been discussed, and Traina or affiliates (for instance Dennis Carlton) have been on an all-out attack on the idea that mark-ups are rising. The problem Traina raises, is one that would have been requested an remedied in the refereeing process anyway. Traina chose to attack this paper to get some of the limelight, but if you look at his paper, it is hard to take his work seriously (if you can't figure out why, look at the updated WP from JdL and JE for an explanation).
Undoubtedly, the way JdL and JE have been responding at conferences is classless, especially given their standing in the field compared to Traina's (and the footnote is a sad sight indeed).
Just note that a Phd student from a top institution chose to write a (very poor) paper attacking a WP in very early stages by some big names. At best Traina is a naive idiot tricked by Carlton and others to attack JdL and JE, at worst..
All in all, disgraceful on both sides. And to the twitterers: stop pretending this doesn't happen in the hard sciences.
-
^ There is plenty of evidence that markups have been rising *some* in the last decade or two, although the pattern is non-monotonic if you extend to earlier decades. (Barkai's much-touted paper showing a steep rise in markups since 1984 goes into reverse as you go backward, and in fact the highest markups on record were achieved in the 1950s.)
The issue is that there is a huge difference between "some rise in markups" and the extraordinary rise that JDL and JE have been pushing, which is from 21% above marginal cost to 61% now. They didn't invent the idea of rising markups, they just came out with a different and more opaque methodology that produced far larger rises than people had gotten using simpler methodologies.
These extreme claims allowed JDL and JE to seize much of the limelight -- and so it definitely seems worth it to me to clarify just how poorly founded these claims are. The weighting issue is by far the most egregious and clearest-cut, and it will be an embarrassment if the QJE doesn't force them to use a proper weighting, but the accounting issue is also a serious weak point and Traina played a major role in pointing that out.
I don't know on what basis you think that Traina was just part of some kind of Chicago Booth, Dennis Carlton-centered conspiracy to suppress discussion of rising markups. It's not like Carlton is one of his primary advisors or anything...?
-
I think the Traina critique is broadly right too, but it is a bit trickier because it gets into this accounting morass that neither JDL and JE nor most macroeconomists know anything about -- what exactly is COGS vs. SG&A, and how does it map onto our notions of variable cost? JDL and JE are certainly wrong to claim that COGS = variable cost, and their results are vulnerable to economically irrelevant accounting changes in the extent to which variable vs. fixed costs have been counted as COGS vs. SG&A over time.
I'm an accountant. It's doubly complicated because we don't necessarily know the macro notion of variable cost. If it's just defined as any cost that varies with output, then both COGS and SG&A have variable and fixed parts.
But if you want to know the difference between COGS and SG&A, COGS (Cost of Goods Sold) is based on the cost of the company's inventories, also known as "product" costs. From a financial accounting perspective, COGS includes manufacturing overhead, which has both variable and fixed components.
SG&A (Selling, General & Administrative) are non-product costs. It includes non-manufacturing overhead, which again has both variable and fixed components.THIS is the key. Thanks bro. In the end, with Compustat data you cannot separate variable and fixed costs (at least for the whole universe of Compustat firms). End of story. However, the new version of JdE & JE, which adds a third co-author, has also estimated markups using census data. They find an increase in markups (though lower than in Compustat). This is consistent with the Barkai story too. Markups, however, are not necessarily a measure of market concentration.
Both Traina and JdL/JE are wrong. And Traina chose to attack first, remember. This indicates that the guy has an attitude. So I don't buy the majority view that JdE/JE are the only bad guys in this story.
-
^ There is plenty of evidence that markups have been rising *some* in the last decade or two, although the pattern is non-monotonic if you extend to earlier decades. (Barkai's much-touted paper showing a steep rise in markups since 1984 goes into reverse as you go backward, and in fact the highest markups on record were achieved in the 1950s.)
The issue is that there is a huge difference between "some rise in markups" and the extraordinary rise that JDL and JE have been pushing, which is from 21% above marginal cost to 61% now. They didn't invent the idea of rising markups, they just came out with a different and more opaque methodology that produced far larger rises than people had gotten using simpler methodologies.
These extreme claims allowed JDL and JE to seize much of the limelight -- and so it definitely seems worth it to me to clarify just how poorly founded these claims are. The weighting issue is by far the most egregious and clearest-cut, and it will be an embarrassment if the QJE doesn't force them to use a proper weighting, but the accounting issue is also a serious weak point and Traina played a major role in pointing that out.
I don't know on what basis you think that Traina was just part of some kind of Chicago Booth, Dennis Carlton-centered conspiracy to suppress discussion of rising markups. It's not like Carlton is one of his primary advisors or anything...?Of course their main markup results are way over the top, just look in their appendices to see that they have selected the juiciest story to present. The Carlton speculation (and nothing more) was just a "best case" speculation of why a grad student would attack a early stages WP from some big names.
-
Both Traina and JdL/JE are wrong. And Traina chose to attack first, remember. This indicates that the guy has an attitude. So I don't buy the majority view that JdE/JE are the only bad guys in this story.
Are you saying that any graduate student who identifies a major methodological issue in a prominent paper and publicizes it is on the "attack" and has an "attitude"?
With that kind of view on the importance of replication and independent work, we are not a science and never will be.
-
Both Traina and JdL/JE are wrong. And Traina chose to attack first, remember. This indicates that the guy has an attitude. So I don't buy the majority view that JdE/JE are the only bad guys in this story.
Are you saying that any graduate student who identifies a major methodological issue in a prominent paper and publicizes it is on the "attack" and has an "attitude"?
With that kind of view on the importance of replication and independent work, we are not a science and never will be.You must not have read the paper by Traina. Plus, thinking of the cost components that JdL and JE initially left out, was always going to be remedied in any serious refereeing process, and probably before that.
-
Both Traina and JdL/JE are wrong. And Traina chose to attack first, remember. This indicates that the guy has an attitude. So I don't buy the majority view that JdE/JE are the only bad guys in this story.
Are you saying that any graduate student who identifies a major methodological issue in a prominent paper and publicizes it is on the "attack" and has an "attitude"?
With that kind of view on the importance of replication and independent work, we are not a science and never will be.No, I am not saying that replication is not important. You cannot infer that from my words. However, the WP of Traina was published despite the concerns of JdL/JE about its arguments and without acknowledging that SG&A could also be treated as a fixed cost.
-
This whole weighting discussion is a waste of time. If your results depend so crucially on how you weight the observations, then they are just not reliable
My guess is that you are thinking about "weighting" in the sense of weighting observations in a regression. That is not so much the relevant question here.
Instead, we are talking about aggregation: how do we appropriately combine population-level micro data into a macro aggregate, given that we know there is lots of heterogeneity at the micro level?
In general, there are more and less theoretically correct ways to do aggregation. JdL and JE's approach, weighting by sales, is horrible and will give estimates that are extremely upward-biased here. Other people, like Baqaee and Farhi and Edmond, Midrigan, and Xu, have pointed out that the right way to measure the aggregate distortion is to weight by costs. JdL and JE are basically just ignoring all these correct arguments because their schtick has been to blow people away with huge numbers, and doing the correct weighting would give much-less-huge numbers.
-
This isn't even that bad. I don't get it. The Traina criticism doesn't make sense to me...their original paper is very clear that you need to only observe one variable cost and its output elasticity. The point is really simple...when you are a firm and optimizing how much of a variable input to use, you think about how much more profit you could get with more of it, and this depends on the markup, the output elasticity and the marginal cost. Marginal cost is directly related to expenditure on the input. Output elasticity is something you estimate separately. This tells you then what production should optimally be with no markups; conversely, the gap between this and what production really is tells you what markups are.
The people saying that "if your argument depends on weighting then it's probably not a good point" lol have you even read the paper? Their weighting scheme is completely reasonable - it's sales weighting. A small firm with no sales should not matter as much in a calculation of rising markups in the macroeconomy as a large firm and this is quite obvious