Poisson PML estimator

[joseph dover]

Hi there!

I would like to do a Poisson PML estimator on STATA but I can't seem to find the command for this?

Thanks!!

Best,

Joseph

[Said Jafar]

Dear Joseph,

Stata build in command for PPML is -poisson- command. But -ppml- command written by Santos Silva is much better.
For (panel data) with fixed effects, consider -xtpoisson- -xtpqml- -poi2hdfe- commands.

Type on stata;

help poisson
ssc install ppml
help ppml
help xtpoisson
ssc install poi2hdfe
help poi2hdfe
ssc install xtpqml
help xtpqml

Also visit to http://personal.lse.ac.uk/tenreyro/LGW.html page.

Best,
Dias

[joseph dover]

Hi Dias!

Thanks a lot for you reply!!

If I am looking at the impact of the introduction of the EURO on the flow of migration (panel data) which one would you recommend I use?

Thanks again!!

Best Regards,

Joseph

[Said Jafar]

Dear Joseph,

I have newly started to use ppml commands and never applied it for migration flows. But it greatly depends on your research question. -ppml- would be better if you do not include fixed effects. xtpoisson and poi2hdfe are better for fixed effects. Pls go over the Log of Gravity page and you will find some more details.

Best,
Dias

[joseph dover]

Thank you Dias!

[dupont john]

Hi Dias!

I am hesitating to use xtreg or xtpoisson

Could I ask you in which situation should I use one and the other?

Thank you very much!

Best Regards,

JD

[Said Jafar]

Dear Dupont,

xtreg is for normal linear regressions. xtpoisson is superior to xtreg for handling zero trade/migration flows. Pls visit the said page for more info on ppml.

Best,
Dias

[joseph dover]

Hi dias!

I just have one last question: I am looking to use ppml fixed effects, is the command for this xtpqml? or xtpoisson?

Thanks!!

Best,


Joseph

[Said Jafar]

Joseph,

This is from -xtpqml help- file:

xtpqml is simply a wrapper for the xtpoisson command that computes robust standard errors.

To my knowledge, this info is outdated as xtpoisson can compute robust standard errors as well, by adding vce(r) option. So, it seems xtpoisson is superior.

All the best

[luka smith]

Hi Dias!

I just have one question for this topic! If I wanted to use PPML Fe, I have to do xtpoisson?

Thanks!

Best,

Luka

[Said Jafar]

Hi Luka.

Not necessarily. poi2hdfe, xi:ppml can also be used. Note that xtpoisson includes pair fixed effects by default.

Best

[Meng Zhang]

Dear Dias,
If I want to use PPML for panel, how to write the command?
Thanks,
Best Regards,

Meng

[Said Jafar]

Dear Meng,

It depends on which code u are using. But in all cases u will need to -xtset- your data. With ppml command, i would be like

ppml depended_var  indep_var1  indep_var2 ...  , cluster(your_panel_var)

Also visit to http://personal.lse.ac.uk/tenreyro/LGW.html page.

Best

[Meng Zhang]

Dear Dias,

Thank you for your reply, I just want to make it sure that if I have written the right code. My denpend_var is in country-sector-year form, there ar 66 countries, and 20 manufacuring industries in 2-digit SIC code from 2006-2014. But my care is only about the China's outwards FDI (OFDI) to 66 countries, and I try to find out the diterminants of China's OFDI. so I write the stata commands as below.

egen id=group(countrycode sic2digit)

xtset id year

ppml ofdi lndist contig lngdp lnchngdp countryskill industryskill countryskillindustryskill countryrdpeople industry_rdintensity countryrdindustryrd bits fta cu economicfreedomindex financedevelopment, cluster(id)

Is it right?

[Said Jafar]

Dear Meng,

The code itself looks right to me.

All the best,
Dias