Fixed Effects Panel Data Quantile Regression

[Devashish Singh]

Hi

I wish to apply quantile regression with fixed effects to a panel dataset of 25 states firms over a 10 year period.

can i apply xtqreg here? i have read that you need a large dataset to apply that, is T=10 large enough? what about qregpd? any help would be appreciated

[Joao Santos Silva]

Dear Devashish Singh,

T=10 is on the short side; I suggest you use the jackknife correction as in this example. The command qregpd does not really estimate a model with fixed effects, at least not in the usual sense.

Best wishes,

Joao

[Devashish Singh]

thank you for your response. i looked up your example. however i had a doubt: what is the significance of c>= 10 in this code

qui bs if c>=10, cluster(idcode) rep(20): xtqreg ln_w age ttl_exp tenure not_smsa south

my c is 9, as opposed to 12 in your case. how should i proceed?

[Devashish Singh]

also professor in your slides you suggested that the bias will be negligible if N/T<10. can i skip the jackknife process as my n/t is 25/9?

[Joao Santos Silva]

Dear Devashish Singh,

The selection of observations with c>=10 is to eliminate some units with very small T, but you can ignore that. I would recommend the JK correction just to be on the safe side but that depends on what you want to do. If this is for a school assignment it is probably OK to ignore it, but if you are giving policy advice the JK correction is the least you should do.

Best wishes,

Joao