Gravity Equation: Panel and PPML

[Gabriel Arrieta]

Good morning Mr. Joao Santos Silva ,

I am doing a panel data model to represent the effects of FTAs (Trade Creation and Trade Diversion) in Trade Flows (Exports and Imports) for Peru and the 90% of its trading partners (from 1995 to 2015). Im using the PPML estimator to estimate the model and also I have been added some year and country-pair dummy variables.

Previously, I estimate my model using FE estimator (according to panel theory), and I obtained some time-invariant variables drop out (as case of distance, language, etc.). So, I would like to ask you why, in the case of PPML estimation, does not drop out the time-invariant variables?

Thanks very much for the atention, Mr. Santos.

[Joao Santos Silva]

Dear Gabriel,

That depends on how you are estimating your model. If you have fixed effects, I recommend that you use -xtpoisson- with clustered standard errors and if you do that all the time invariant variables will drop.

Best wishes,

Joao

[Gabriel Arrieta]

Dear Prof. Santos Silva,

Thank you very much for your answer.

And I have some related questions:

- Can I consider this estimation method (xtpoisson) as a special case of PPML estimator (for panel fixed effects PPML estimation)? Or are both different estimators?

- If I am using year and country-pair dummy variables, Should be considered (these dummy variables) as fixed effects modelling?

- If I consider to drop out with my FE estimation, all the time invariant variables? How can handle with my "distance" variable that is very important for my Gravity equation analysis? Should I use Random Effects to obtain an estimated coefficient for distance?

All the best,
Gabriel

[Joao Santos Silva]

Dear Gabriel,

1) xtppml is indeed a PPML estimator

2) Indeed, people now call fixed effects to those dummies

3) Do not use random effects; if you use pair-fixed effects the coefficient of distance cannot be identified. Many models do not include pair-fixed effects.

Best wishes,

Joao

[Gabriel Arrieta]

Dear Prof. Santos Silva,

So, can I estimate my model using xtppml, that is the same "xtpoisson" or "xtpqml"?

Thank you for your answers and your helpful comments professor.

Best regards,
Gabriel

[Joao Santos Silva]

Sorry, I meant xtpoisson; xtppml does not exist (AFAIK). xtpqml should give the same results as xtpoisson.

[Gabriel Arrieta]

Dear Prof. Santos Silva,

Thank you very much for your helpful comments and suggestions.

Best regards,
Gabriel

[Gabriel Arrieta]

Sorry, I meant xtpoisson; xtppml does not exist (AFAIK). xtpqml should give the same results as xtpoisson.

Dear Prof. Santos Silva,

I just notice that this xtpoisson estimator does not report R-squared value. How can I obtain it?

Thank you very much.

Best regards,
Gabriel

[Joao Santos Silva]

Here is an example of how to do it

clear all
sysuse auto
xtset rep78
xtpoisson price mpg , fe
predict double xb if e(sample)==1
g double yhat=exp(xb)
egen meany=mean(price) if e(sample)==1, by(rep78)
egen meanyhat=mean(yhat) , by(rep78)
gen double exp_alpha=meany/meanyhat  
replace yhat=yhat*exp_alpha
qui corr yhat price
di "R2 = " r(rho)^2

All the best,

Joao

[Merve Kadiroglu]

Hello,

I am using PPML to estimate the determinants of export.
My ppml estimation commands are;


ppml exports_son lgdp_reporter lgdp_partner lpop_reporter lpop_partner lland_reporter lland_partner ldist colony contig comlang_off , cluster(pairid)

ppml exports_son lgdp_reporter lgdp_partner lpop_reporter lpop_partner lland_reporter lland_partner ldist colony contig comlang_off year_*, cluster(pairid)

ppml exports_son lgdp_reporter lgdp_partner lpop_reporter lpop_partner lland_reporter lland_partner ldist colony contig comlang_off year_* reporter_* partner_*,cluster(pairid)

xtpqml exports_son lgdp_reporter lgdp_partner lpop_reporter lpop_partner lland_reporter lland_partner ldist colony contig comlang_off year_*, fe cluster(pairid)


I have 2 questions,

1)Are the commands correct for ppml est.?

2) And second i want to get the predicted value of exports. But i don know which stata command /commands i need to use for prediction.

I will be very glad if somebodys answering my questions.

Best

[Joao Santos Silva]

Dear Merve,

1) The commands you give estimate models with different sets of fixed effects and they look correct (if those are the regressors and fixed effects you want to include).

2) With ppml you can get predictions using the predict command after the estimation. With xtpqml things are a bit more complicated, but you can get them as in #9 above (the predicted values are in yhat in the penultimate line).

Best wishes,

Joao

[Merve Kadiroglu]

Dear Joao,

Thank you very much for your answer.

I used firstly "ppml exports lgdp_reporter lgdp_partner lpop_reporter lpop_partner lland_reporter lland_partner ldist colony contig comlang_off year_* reporter_* partner_*,cluster(pairid)" and than "predict mu" commands. And a mu column appeared showed like below.

As i understand for ex. " In 2001 Predicted value of export is 2901813 from Tur. to Aut. " Is it correct or not?

Thanks for instance,

Best Wishes.

year	reporter	partner	mu
2001	TUR	AUT	2901813
2002	TUR	AUT	3313635
2003	TUR	AUT	3952249
2004	TUR	AUT	5004043
2005	TUR	AUT	6210501
2006	TUR	AUT	7298055
2007	TUR	AUT	8382143
2008	TUR	AUT	8662180
2009	TUR	AUT	9155061
2010	TUR	AUT	8143444
2011	TUR	AUT	9238369
2012	TUR	AUT	1.03E+07
2013	TUR	AUT	1.06E+07
2014	TUR	AUT	1.13E+07
2015	TUR	AUT	1.13E+07
2016	TUR	AUT	1.14E+07

[Joao Santos Silva]

That is correct.

[Merve Kadiroglu]

Thank you very much for your helpful comments..

Best Regards

[Javier Rua Montes]

Dear Joao,

Thank you very much for your answer.

I used firstly "ppml exports lgdp_reporter lgdp_partner lpop_reporter lpop_partner lland_reporter lland_partner ldist colony contig comlang_off year_* reporter_* partner_*,cluster(pairid)" and than "predict mu" commands. And a mu column appeared showed like below.

As i understand for ex. " In 2001 Predicted value of export is 2901813 from Tur. to Aut. " Is it correct or not?

Thanks for instance,

Best Wishes.

year	reporter	partner	mu
2001	TUR	AUT	2901813
2002	TUR	AUT	3313635
2003	TUR	AUT	3952249
2004	TUR	AUT	5004043
2005	TUR	AUT	6210501
2006	TUR	AUT	7298055
2007	TUR	AUT	8382143
2008	TUR	AUT	8662180
2009	TUR	AUT	9155061
2010	TUR	AUT	8143444
2011	TUR	AUT	9238369
2012	TUR	AUT	1.03E+07
2013	TUR	AUT	1.06E+07
2014	TUR	AUT	1.13E+07
2015	TUR	AUT	1.13E+07
2016	TUR	AUT	1.14E+07

Dear Mr. Joao

My question is about the possibility of predicting values out of sample data with a PPML model. Is this possible? Is there a command to execute in Stata?