moderation with quantile regression

[sedki zn]

Dear all

i hope you are doing good

i'm working on my research paper and i want to try the moderated effect of firm performance (ROA) in the relationship between CEO's compensation and strategic risk taking (R&D) using quantile regression approach

i have tried to add the interaction variable ( c.stockoptions##c.ROA ) to the regression but i have got the following error msg " factor variables and time-series operators not allowed"

i hope there is a way to run the quantile regression with moderation.
i would appreciate your help

Best regards
Sedki

[sedki zn]

Dear Joao Santos Silva

i would appreciate your help, as you do always

kind regards

[FernandoRios]

Hi Sedki
you didn’t mention which command you are using for the estimation of qreg, thus it’s difficult to say why you are getting that error
f

[sedki zn]

Dear FernandoRios
thank you for your response

yes sorry, you are right

actually i am using the qregpd command

[Joro Kolev]

Generate the interaction manually.

gen optionsXroa = stockoptions*roa

and then use it as a variable in the regression.

[Joao Santos Silva]

Dear sedki zn,

Mane sure you really want to use qregpd; it does not estimate a model with fixed effects as you may think.

Best wishes,

Joao

[sedki zn]

Dear Joao Santos Silva

Actually im suspecting the issue of endogeneity, that's why im using the qregpd with IV
if the qregpd doesn't deal with a model with fixed effect, which is the appropriated command ?

best regards

[FernandoRios]

Dear sedki zn
It isnt the case that qregpd doesn't deal with FE, but that it accounts for them in a way that is different from what you would do with standard fixed effects (think for example of xtreg, fe).
If you read Powell's paper, you may be able to understand how is it that FE are accounted for with this approach.
Regarding your second question. I don't think an "appropriate" command exists yet.
-xtqreg- deals with fixed effects in the more traditional way, imposing some structural assumptions on the heterogeneity, but does not deal with endogeneity. and -ivqreg2- can estimate models with endogeneity, but does not handle large number of fixed effects.

Perhaps you can try a Control function approach, combined with xtqreg and bootstrap, to get something closer to what you need.
HTH
F

[sedki zn]

Dear FernandoRios

thank you very much for your response

if i got your answer right, the qregpd can deal with standard fixed effects ? ( not year or industry fixed effect) and in this case i can use it ?
and if i need to add "specific" fixed effects ( as year fixed effect or industry fixed effect), which is the maximum number that ivqreg2 can handle ? ( 1, 2 ... ?)
and does ivqreg2 work for panel data??

Best regards
SEDKI