Close

Complete Captcha

• 1 year - 34 posts - Latest - RSS
• Thread: 37 Goods vs 16 No Goods

1. Economist
   cb31
   

   DLW (2012) became the de facto method to "compute" market power.

   This lecture (https://www.nuffield.ox.ac.uk/media/3209/mktpower.pdf) by Steve Bond (future Nobel Prize winner) just destroys the method. I quote:

   "In the benchmark case of Cobb Douglas technology and constant elasticity
   demand, there is no information about the price-cost markup in the ratio of
   an estimated revenue elasticity to the corresponding cost share for a perfectly
   flexible input"

   "More generally, we should be cautious in inferring anything about markups
   or market power from the ratio of an estimated revenue elasticity to the
   corresponding cost share"

   1 year ago # QUOTE 2 Jab 2 No Jab !

2. Economist
   e615
   

   true for cobb-douglas, but what about translog? That is where most of the action happens otherwise markups differ only due to differences in cost-shares between firms..

   1 year ago # QUOTE 0 Jab 0 No Jab !

3. Economist
   cb31
   

   true for cobb-douglas, but what about translog? That is where most of the action happens otherwise markups differ only due to differences in cost-shares between firms..

   As the slides state at the end, translog stuff is still work in progress. There is probably some value there. In any case, most empirical work uses Cobb-Douglas.

   1 year ago # QUOTE 0 Jab 2 No Jab !

4. Economist
   45f1
   

   The revenue data point is old news, and doesn't matter in practice.

   The bigger deal is the non identification, check out some of the graphs in Flynn, Gandhi, & Traina: See https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3358472

   1 year ago # QUOTE 0 Jab 1 No Jab !

5. Economist
   a0c9
   

   DLW (2012) became the de facto method to "compute" market power.
   This lecture (https://www.nuffield.ox.ac.uk/media/3209/mktpower.pdf) by Steve Bond (future Nobel Prize winner) just destroys the method. I quote:
   "In the benchmark case of Cobb Douglas technology and constant elasticity
   demand, there is no information about the price-cost markup in the ratio of
   an estimated revenue elasticity to the corresponding cost share for a perfectly
   flexible input"
   "More generally, we should be cautious in inferring anything about markups
   or market power from the ratio of an estimated revenue elasticity to the
   corresponding cost share"

   But DW assumes they observe physical output, not just revenue

   1 year ago # QUOTE 0 Jab 1 No Jab !

6. Economist
   9cfe
   

   Classical singers and art directors also need attentions.

   1 year ago # QUOTE 0 Jab 0 No Jab !

7. Economist
   7db9
   

   maybe in the next version there would be another footnote 1

   1 year ago # QUOTE 0 Jab 0 No Jab !

8. Economist
   ce34
   

   What does footnote 1 mean?

   1 year ago # QUOTE 0 Jab 0 No Jab !

9. Economist
   9c1a
   

   DLW (2012) became the de facto method to "compute" market power.
   This lecture (https://www.nuffield.ox.ac.uk/media/3209/mktpower.pdf) by Steve Bond (future Nobel Prize winner) just destroys the method. I quote:
   "In the benchmark case of Cobb Douglas technology and constant elasticity
   demand, there is no information about the price-cost markup in the ratio of
   an estimated revenue elasticity to the corresponding cost share for a perfectly
   flexible input"
   "More generally, we should be cautious in inferring anything about markups
   or market power from the ratio of an estimated revenue elasticity to the
   corresponding cost share"

   But DW assumes they observe physical output, not just revenue

   read the paper before posting nonsense!

   1 year ago # QUOTE 0 Jab 1 No Jab !

10. Economist
    a0c9
    

    I did. They make assumptions such that they "observe" physical output. Read appendix of the Flynn Gandhi Traina paper for the grim details as they do it too.

    1 year ago # QUOTE 0 Jab 0 No Jab !

11. Economist
    d553
    

    I think you are wise ^

    1 year ago # QUOTE 0 Jab 0 No Jab !

12. Economist
    b61f
    

    high probability that OP is steve bond. lol @ future nobel winner. Dude isn't even a professor.

    1 year ago # QUOTE 4 Jab 1 No Jab !

13. Economist
    34fb
    

    If someone is HRM and 20 years out, then isn't it natural to have 20 AERs?

    1 year ago # QUOTE 0 Jab 3 No Jab !

14. Economist
    9c1a
    

    I did. They make assumptions such that they "observe" physical output. Read appendix of the Flynn Gandhi Traina paper for the grim details as they do it too.

    No James, DW(2012) shows that levels (markups) are not measured consistently when revenue-productivity is used, but changes in markups are.

    1 year ago # QUOTE 2 Jab 0 No Jab !

15. Economist
    b61f
    

    If someone is HRM and 20 years out, then isn't it natural to have 20 AERs?

    Idk about hrm. Oxford is more of a lrm.

    1 year ago # QUOTE 0 Jab 1 No Jab !

16. Economist
    e7dd
    

    If someone is HRM and 20 years out, then isn't it natural to have 20 AERs?

    Better to have the most cited paper in economics than a load of top 5s that nobody reads surely?

    1 year ago # QUOTE 2 Jab 0 No Jab !

17. Economist
    cf78
    

    I did. They make assumptions such that they "observe" physical output. Read appendix of the Flynn Gandhi Traina paper for the grim details as they do it too.

    The deflated sales data problem is very old: https://doi.org/10.1002/(SICI)1099-1255(199607)11:4%3C343::AID-JAE404%3E3.0.CO;2-4

    And it doesn't usually matter quantitatively: https://doi.org/10.1111/j.1467-9442.2005.00431.x

    Ideally, you want firm-specific same-quality q. Instead, you have firm-level pq and industry-level p.

    The bias comes from the within-industry price dispersion. If a firm has a higher price than average, then when you deflate, it'll have an incorrectly high q, which creates a positive bias in the elasticities and mark-ups. It works vice versa for a firm with a lower price than average, so at the industry level, it averages out.

    Not only that, but if the price differences are about quality, then you actually WANT to use the deflator method. You need to get q to be comparable across firms, i.e. quality-adjusted. It's the same idea as in labour: if you have a worker that's twice as productive and gets paid twice as much because of it, then you want to count her as two workers in the production function estimation.

    The Flynn, Gandhi, and Traina identification problem is very different. "When the flexible input expenditure is high, we cannot disentangle whether it is because its output elasticity is high, or because productivity is high."

    And unlike the deflated sales problem, this problem DOES matter quantitatively.

    1 year ago # QUOTE 2 Jab 0 No Jab !

18. Economist
    f36c
    

    Reminder: production function estimation is just total BS.

    1 year ago # QUOTE 9 Jab 1 No Jab !

19. Economist
    7f8a
    

    Reminder: production function estimation is just total BS.

    true?

    1 year ago # QUOTE 0 Jab 0 No Jab !

20. Economist
    fb13
    

    confirmed.

    1 year ago # QUOTE 0 Jab 0 No Jab !

Reply »