Is This the Final Tomb Stone on De Loecker's Markups?
-
Interesting. What are the other tombstones?
https://www.nber.org/papers/w24800
And discussion here (among other drama): https://www.econjobrumors.com/topic/de-loecker-and-eckhout-go-on-all-out-attack
-
I think Jaumandreu has a very valid point.
Differently from the Traina's debate, it is not about aggregation, it is about how you actually measure markups at the firm level.So, it actually hits the IO DE Loecker (the one that was considered _good_), not the Macro De Loecker (the one that is highly debated, to say the least)
-
Have they adressed the weighting mistake yet?
Briefly on page 11: http://www.janeeckhout.com/wp-content/uploads/Thoughts.pdf
Maybe some will be convinced by it. I'm more convinced by Edmond, Midrigan, and Xu (2018), and especially by 9329 here: https://www.econjobrumors.com/topic/de-loecker-and-eckhout-go-on-all-out-attack/page/4
I think the intuition for the general theoretical result is non-obvious, so let's cover one place where the intuition is obvious: markups over total costs. (Btw, this comment keeps getting deleted, so I'm trying again.)
Imagine an economy where 10% of sales are done by firms who have a markup of 10x over total costs, and 90% of sales are done by firms who have a markup of 1x over total costs.
If we do a sales-weighted average of markups, we will get an aggregate markup of 10%*10 + 90%*1 = 1.9.
If we do a cost-weighted average of markups, then we will get an aggregate markup of 1/91*10 + 90/91*1 = 1.1.
Which is right, the sales-weighted average 1.9 or the cost-weighted average 1.1? Well, if we just took aggregate*sales for the entire economy and divided them by aggregate costs, we'd get the latter, 1.1. So if we want a measure of the total gap between sales and costs for the entire economy, we want to aggregate using cost weights. This is one instance of the general point that to recover an aggregate ratio, you need to weight the individual ratios by their denominators. (And it makes a very, very big difference, as we can see in this example.)
-
jaumandreu paper is misleading, if markups are the same across firms in a given time period or industry, DW is fine.
I guess you meant demand, markups are exactly what you want to measure idiosyncratically.
But then you should measure elasticities (I. E. Production functions) within each cell that may face similar demand.
This not only implies industry-And-period, but also age, as we know that young firms face different demand than older ones (Haltiwanger, Syverson,... Moreira... Etc.) -
They seem to miss the point about weighting or are being deliberately misleading. The argument for cost weighting doesn’t depend at all on CES assumptions, as the example below shows.
Have they adressed the weighting mistake yet?
Briefly on page 11: http://www.janeeckhout.com/wp-content/uploads/Thoughts.pdf
Maybe some will be convinced by it. I'm more convinced by Edmond, Midrigan, and Xu (2018), and especially by 9329 here: https://www.econjobrumors.com/topic/de-loecker-and-eckhout-go-on-all-out-attack/page/4I think the intuition for the general theoretical result is non-obvious, so let's cover one place where the intuition is obvious: markups over total costs. (Btw, this comment keeps getting deleted, so I'm trying again.)
Imagine an economy where 10% of sales are done by firms who have a markup of 10x over total costs, and 90% of sales are done by firms who have a markup of 1x over total costs.
If we do a sales-weighted average of markups, we will get an aggregate markup of 10%*10 + 90%*1 = 1.9.
If we do a cost-weighted average of markups, then we will get an aggregate markup of 1/91*10 + 90/91*1 = 1.1.
Which is right, the sales-weighted average 1.9 or the cost-weighted average 1.1? Well, if we just took aggregate*sales for the entire economy and divided them by aggregate costs, we'd get the latter, 1.1. So if we want a measure of the total gap between sales and costs for the entire economy, we want to aggregate using cost weights. This is one instance of the general point that to recover an aggregate ratio, you need to weight the individual ratios by their denominators. (And it makes a very, very big difference, as we can see in this example.)
-
jaumandreu paper is misleading, if markups are the same across firms in a given time period or industry, DW is fine.
I guess you meant demand, markups are exactly what you want to measure idiosyncratically.
Yes, thanks for clarifying
Why would we assume demand is the same across firms in a given time period? Seems pretty unrealistic to me...
-
jaumandreu paper is misleading, if markups are the same across firms in a given time period or industry, DW is fine.
I guess you meant demand, markups are exactly what you want to measure idiosyncratically.
Yes, thanks for clarifying
Why would we assume demand is the same across firms in a given time period? Seems pretty unrealistic to me...
perfect competition, symmetric monopolistic competition, for example. it is done all the time in this literature.
-
perfect competition, symmetric monopolistic competition, for example. it is done all the time in this literature.
I understand that. But these are simplifying assumptions that make the modeling easier. The question is are these assumptions good enough to represent the real world? My prior on this is no, but I'm open to being shown that I'm wrong.