[Joao Santos Silva]

Dear muhammad akhtaruzzaman,

You can do it either way but in general the results will answer different questions. In the first case you will learn only about NZ's outward FDI whereas in the in the second case you will learn about outward FDI in general.

Best wishes,

Joao

[Javier Rua Montes]

Regards Professor Santos silva

According to what has been read, is it correct to make the following interpretation?

If the model is:
import (tonnes) = 0.085 + 0.52 * PBIimp + 0.35 * PBIexp -0.25 * Dist

The interpretation of the coefficients is:

If the importer's GDP increases one unit, does the volume of import between two countries increase by 52% or 0.52 Tonnes?

Thanks for your responses!

Javier

[Malela Oscar]

Hi everyone, I'm new to gravity model but I'm learning hard now, I want to analyze China’s Belt and Road Initiative and its Impacts on the Trade Patterns of Tanzania,Here is my data set link https://drive.google.com/file/d/1xC0...ew?usp=sharing Would anybody check my data and advice me please!! I want to know if the data set is correct and or what is the format for my dataset should be?

[Joao Santos Silva]

Dear Javier Rua Montes,

Is your model really linear?

Best wishes,

Joao
PS: Malela Oscar, it would be better to start a new thread for your question.

[Maryam Zia]

Dear Professor Silva,

I have few questions:

1. In the paper you said that PPML takes care of zeros in dependent variable naturally but I couldn't get any further explanation for that in paper. Please let me know if I can find something on it in more detail.
2. Second you also said that mean is proportional to variance, I want to understand what does it mean and is it relevant to taking care of zeros in dependent variable. I also read the paper ,"Testing non-nested models for non-negative data with many zeros", but still can't understand fully what is the explanation for taking care of Zeros

Sorry for my naive questions

Best,
Maryam Zia

[Joao Santos Silva]

Dear Maryam Zia,

1 - Because the model is estimated in levels, the zeros are just like any other value so there is no need to worry about them.
2 - PPML assumes that the conditional variance is proportional to the conditional mean. However, the estimator is still valid (and generally well behaved) even if that assumption is not valid. This has noting to do with the zeros.

Best wishes,

Joao

[Maria Lopez]

Dear professor Jao,

I am runnig a gravity model for Agricultural Exports in Nicaragua with 27 trade partners. I couldn't find any literature about how the sign of the remotenesss variable should be. I am using remotoness based on (Head 2003) for exporter and importers.

Thank you.

Maria

[Joao Santos Silva]

Dear Maria Lopez,

In the Log of Gravity, our PPML estimates of the remoteness are positive, but I think you will find other papers commenting on the effect of that variable. Note, however, that nowadays that variable is not often used because the model include origin and destination dummies.

Best wishes,

Joao

[Maria Lopez]

Dear Professor Jao,

Thank you so much for your answer.

For being more specific:

My independent variables for the base model are:

1. Distance
2. Contiguity
3. Common Language
4. Same Country
5. GDP current US for the importer and Exporter
6. TD: Time dummy

and F_* was created like this
xi, pre(F_1) i.Importer

- In the first estimation I am running as depedent variable the natural log of the agriultural exports of Nicaragua to its 27 trade partners. This variables include the zeros.

reg ln_AgriExport ln_distcap_ contig_ comlang_off_ smctry_ lnGDPImporter lnGDPNIC F_* TD*, robust

-In this way I am trying to work with zeros and assuming at least 1 dollar exported.

reg ln_Exports_D1 ln_distcap_ contig_ comlang_off_ lnGDPImporter lnGDPNIC F_* TD*, robust

- And base on the Practical Guide to trade policy Analysis of the WTO, I use agricultural expoort in levels.

ppml AgriExport ln_distcap_ contig_ comlang_off_ smctry_ lnGDPImporter lnGDPNIC F_* TD*

Do you have any advice ?

I really appreciate your time and support.

Best wishes,

Maria

[Joao Santos Silva]

Dear Maria,

My advice is to stick to the estimation in levels and ignore the ones based on the logs. Apart from that, just follow the advice on the WTO book.

Best wishes,

Joao

[Maria Lopez]

Dear proffesor Jao,

Thank you so much for your advice.

Best wishes,

Maria

[Maria Lopez]

Hello!!

Does anyone know how to run ppml with random effect in stata?

Is it possible?

Thank you

Maria

[Joao Santos Silva]

Do not run Poisson regression with random effects; it is not a good idea.

Best wishes,

Joao

[Maria Lopez]

Thank you for your answer professor Jao.

Best wishes,

Maria

[Maria Lopez]

Dear Professor Jao,

I have a question regarding how to interpret the coefficients of my gravity equation, using PPML estimator . Since my dependent variable is in level (agricultural exports) and the independent variables such as, GDP and Distance are in log. How can I explain the effect of these variables over the agricultural exports.

Dependent Variable	PPML
Log Distance	-1.064***
	(0.06)
Contiguity	-0.834***
	(0.13)
Common Language	0.0248
	(0.20)
Same Country	1.078***
	(0.20)
Log Importer's GDP	0.739***
	(0.03)
Log Nicaragua's GDP	0.878***
	(0.18)
Remoteness_Importer
Remoteness_NIC
Constant	8.170***
	-1.771
Observations	648
R-squared	0.825
Time Dummy	Yes

shoul I apply this formula: (coeffcient/100), for the importer's GDP and Distance?

Thank you so much for your time.

Maria