Close

Complete Captcha

• 1 year - 156 posts - Latest - RSS
• Thread: 268 Goods vs 94 No Goods

1. Economist
   725b
   

   I dont get the fuss. With each paper there is an N-dimensional matrix of research design choices that fall within reasonability bounds (usually meaning design choice is defensible by citing prior research). FHK present M cells of this matrix. The cells they present suggest an effect exists. BB et al present J different cells of the matrix. The cells they present suggest the effect does not exist. Both present results for cells within the matrix of reasonable design choices. We learn its not as clear cut as FHK would lead us to believe. Done.
   But the choices made by BB et al (J cells) are not ex ante more credible than the choices made by FHK (M cells)

   Ideally we would like to know the distribution of results across this matrix.

   2 months ago # QUOTE 0 Jab 0 No Jab !

2. Economist
   f20b
   

   I feel FHK is worried now.

   Why only FHK? All four papers failed to replicate

   2 months ago # QUOTE 3 Jab 0 No Jab !

3. Economist
   0748
   

   I dont get the fuss. With each paper there is an N-dimensional matrix of research design choices that fall within reasonability bounds (usually meaning design choice is defensible by citing prior research). FHK present M cells of this matrix. The cells they present suggest an effect exists. BB et al present J different cells of the matrix. The cells they present suggest the effect does not exist. Both present results for cells within the matrix of reasonable design choices. We learn its not as clear cut as FHK would lead us to believe. Done.
   But the choices made by BB et al (J cells) are not ex ante more credible than the choices made by FHK (M cells)

   This is ridiculous. If you are arguing that barriers to shorting affect executive compensation or analyst incentives, why should these only matter if you choose the sample of affected firms a certain way?

   2 months ago # QUOTE 0 Jab 0 No Jab !

4. Economist
   457f
   

   bump

   2 months ago # QUOTE 0 Jab 0 No Jab !

5. Economist
   caa4
   

   any Reg SHO working paper

   1 month ago # QUOTE 0 Jab 0 No Jab !

6. Economist
   c2b9
   

   Every Reg SHO paper showing an effect of Reg SHO other than a micro-structure effect is p-hacked garbage. Plain and simple. Ask an actual manager whether a temporary, random, short-sale ban affects corporate policies and you'd get either laughter or a quizzical look.

   1 month ago # QUOTE 5 Jab 0 No Jab !

7. Economist
   499b
   

   Great suggestion. Let us just ask managers for answers of CF questions and stop running regressions.

   Every Reg SHO paper showing an effect of Reg SHO other than a micro-structure effect is p-hacked garbage. Plain and simple. Ask an actual manager whether a temporary, random, short-sale ban affects corporate policies and you'd get either laughter or a quizzical look.

   1 month ago # QUOTE 1 Jab 0 No Jab !

8. Economist
   d1b9
   

   My impression is that, what managers think as important might not turn out to be important in regressions.
   But what managers regard as trivial is almost always indeed irrelevant in regressions..

   Great suggestion. Let us just ask managers for answers of CF questions and stop running regressions.

   Every Reg SHO paper showing an effect of Reg SHO other than a micro-structure effect is p-hacked garbage. Plain and simple. Ask an actual manager whether a temporary, random, short-sale ban affects corporate policies and you'd get either laughter or a quizzical look.
   

   1 month ago # QUOTE 5 Jab 0 No Jab !

9. Economist
   73a7
   

   Great suggestion. Let us just ask managers for answers of CF questions and stop running regressions.

   Every Reg SHO paper showing an effect of Reg SHO other than a micro-structure effect is p-hacked garbage. Plain and simple. Ask an actual manager whether a temporary, random, short-sale ban affects corporate policies and you'd get either laughter or a quizzical look.
   

   Many papers using Reg SHO rely on managers being aware of the change and responding optimally to it. In the real world, most managers never even heard of it.

   1 month ago # QUOTE 4 Jab 0 No Jab !

10. Economist
    b0be
    

    I dont get the fuss. With each paper there is an N-dimensional matrix of research design choices that fall within reasonability bounds (usually meaning design choice is defensible by citing prior research). FHK present M cells of this matrix. The cells they present suggest an effect exists. BB et al present J different cells of the matrix. The cells they present suggest the effect does not exist. Both present results for cells within the matrix of reasonable design choices. We learn its not as clear cut as FHK would lead us to believe. Done.
    But the choices made by BB et al (J cells) are not ex ante more credible than the choices made by FHK (M cells)

    This is ridiculous. If you are arguing that barriers to shorting affect executive compensation or analyst incentives, why should these only matter if you choose the sample of affected firms a certain way?

    You missed the point. We dont know truth. We are trying to discover it. If I show you 10 specifications and they all hold up, great. If they only held up under 5... does that somehow reduce the amount of information I have provided? Somehow reduce the amount you learn?

    Different inferences sure, but we learn from both FHK and BB et al...

    In this case, ppl overweighted FHK, BB et al are saying place less weight on the info in their study. NOT place ZERO weight on the study.

    - Bayesian

    1 month ago # QUOTE 1 Jab 4 No Jab !

11. Economist
    f537
    

    Don't bring your logic and reasonableness here.

    I dont get the fuss. With each paper there is an N-dimensional matrix of research design choices that fall within reasonability bounds (usually meaning design choice is defensible by citing prior research). FHK present M cells of this matrix. The cells they present suggest an effect exists. BB et al present J different cells of the matrix. The cells they present suggest the effect does not exist. Both present results for cells within the matrix of reasonable design choices. We learn its not as clear cut as FHK would lead us to believe. Done.
    But the choices made by BB et al (J cells) are not ex ante more credible than the choices made by FHK (M cells)

    This is ridiculous. If you are arguing that barriers to shorting affect executive compensation or analyst incentives, why should these only matter if you choose the sample of affected firms a certain way?

    You missed the point. We dont know truth. We are trying to discover it. If I show you 10 specifications and they all hold up, great. If they only held up under 5... does that somehow reduce the amount of information I have provided? Somehow reduce the amount you learn?
    Different inferences sure, but we learn from both FHK and BB et al...
    In this case, ppl overweighted FHK, BB et al are saying place less weight on the info in their study. NOT place ZERO weight on the study.
    - Bayesian

    1 month ago # QUOTE 1 Jab 2 No Jab !

12. Economist
    aaa6
    

    If I show you 10 specifications and they all hold up, great.

    Have you ever read anything on Gelman's blog?

    1 month ago # QUOTE 0 Jab 0 No Jab !

13. Economist
    b0be
    

    The way to deal with the garden of forking paths, is to show the paths. Show all 10 specifications. Not sure what the scientific argument is against that. (of course publication incentives, but that is what we are trying to combat)

    If I show you 10 specifications and they all hold up, great.

    Have you ever read anything on Gelman's blog?

    1 month ago # QUOTE 1 Jab 0 No Jab !

14. Economist
    717b
    

    Glad to see people debunking the crazy "Eliminating Reg SHO ruined the world" nonsense.

    Also, if killing Reg SHO did reduce earnings management, that is probably a good thing. If that leads to reduced investment (in *those firms*). then good because investors are getting a clearer picture of the risk and will allocate their capital to better firms.

    1 month ago # QUOTE 1 Jab 0 No Jab !

15. Economist
    43ad
    

    But they behave AS IF they did

    Great suggestion. Let us just ask managers for answers of CF questions and stop running regressions.

    Every Reg SHO paper showing an effect of Reg SHO other than a micro-structure effect is p-hacked garbage. Plain and simple. Ask an actual manager whether a temporary, random, short-sale ban affects corporate policies and you'd get either laughter or a quizzical look.
    

    Many papers using Reg SHO rely on managers being aware of the change and responding optimally to it. In the real world, most managers never even heard of it.

    1 month ago # QUOTE 0 Jab 0 No Jab !

16. Economist
    f944
    

    I dont get the fuss. With each paper there is an N-dimensional matrix of research design choices that fall within reasonability bounds (usually meaning design choice is defensible by citing prior research). FHK present M cells of this matrix. The cells they present suggest an effect exists. BB et al present J different cells of the matrix. The cells they present suggest the effect does not exist. Both present results for cells within the matrix of reasonable design choices. We learn its not as clear cut as FHK would lead us to believe. Done.
    But the choices made by BB et al (J cells) are not ex ante more credible than the choices made by FHK (M cells)

    This is ridiculous. If you are arguing that barriers to shorting affect executive compensation or analyst incentives, why should these only matter if you choose the sample of affected firms a certain way?

    You missed the point. We dont know truth. We are trying to discover it. If I show you 10 specifications and they all hold up, great. If they only held up under 5... does that somehow reduce the amount of information I have provided? Somehow reduce the amount you learn?
    Different inferences sure, but we learn from both FHK and BB et al...
    In this case, ppl overweighted FHK, BB et al are saying place less weight on the info in their study. NOT place ZERO weight on the study.
    - Bayesian

    This would be great if the paper was framed as "we show evidence of x under specification choices y and no evidence otherwise" instead of "we conclude that executives do z because of our evidence of x"

    1 month ago # QUOTE 1 Jab 0 No Jab !

Reply

Post <blockquote>lol, you seem quite dumβ, especially because you are too lazy to actually read AER's data policy. Luckily, Black et al. have footnote 41 for people like you: "The American Economic Review has a general requirement to post data and code, but authors are not required to post data from commercial sources, such as Compustat. LLS posted their regression code, but not their sample, nor the code used to obtain their data from commercial sources. The posted code does not indicate the sample selection choices they made, and their paper contains very few sample selection details." <blockquote>Well I don’t read or care about JF papers. An AER paper not replicable? I am listening. </blockquote></blockquote>

Send Post »

Markup: a blockquote code em strong ul ol li.