using the distance on Gravity model

[Raisa Yuliana]

Dear STATALIST,
I use the basic gravity model with OLS using STATA 14 for my econometry analysis. But, the OLS result for distance always omitted, i already tried it so many times from change the distance become dependent variable until amend the data for distance itself and the result still omitted. I don't know what is going on with my model since i use the basic model. Can somebody explain to me what is going on with this formula?
Thank you

[Joao Santos Silva]

Raisa,

We can only help you if you show us what results you get, but my guess is that you are including something that is collinear with distance.

Best wishes,

Joao

[Raisa Yuliana]

I don't know..since i just follow the basic model for gravity, which is E=GDPi+GDPj+D, this is my result for OLS, but when i test the omitted variable with Ramsey Test the p>value is higher than 0.05 which means no omitted variable. The ACFTA is my dummy for my model since I want to analyze the impact of ACFTA on Indonesia's Export to China.
Thank you

[Joao Santos Silva]

Raisa,

If you are only looking at exports from Indonesia to China, then distance is constant and therefore collinear with the constant. Also, note that Ramsey's test is not a test for omitted variables.

Best wishes,

Joao

[Raisa Yuliana]

So, you mean I can not use this model for my analysis or I don't need the distance in my case? Ramsey's test not a test for omitted variables?? i don't know about that since I saw the analyze step use STATA from my friend. Sorry, if I ask too many questions since I'm new with all of this.
Thank you

[Joao Santos Silva]

Indeed, with your data distance is not needed. Ramsey's test is a test for the correct specification of the model, conditional on the variables you have. So it cannot detect that soem othe variable is omitted. Unfortunately there are lots of misconceptions about this test.

Joao

[Raisa Yuliana]

I see. What do you think if I use the Population of both country instead of the distance? Do you it can represent the distance? *_*
-Raisa-

[Raisa Yuliana]

Joao, I am still curious about the distance. Why the distance is constant because I "only looking the export from Indonesia to China"?Is it because the wrong method of my analysis or the data that i use?
Thank you,
-Raisa-

[Joao Santos Silva]

Raisa,

I suggest you compute descriptive statistics for distance to check that it is actually constant, but my guess is that you have time series data for exports from Indonesia to China and in that case the distance will be constant because the distance between Indonesia and China does not change over time (at least not in a way that is significant for your work).

Population will not represent distance; if distance is constant, just ignore it.

Best wishes,

Joao

[Raisa Yuliana]

Well, your guess is right. I used time series data for exports from Indonesia to China.
Ok, I understand right now about this distance problem. In another word, since I just use the export from one country to another country and the distance between this two nations will be same all the time that makes the distance will be constant. Instead, if I use the export from Indonesia to some countries, let's say China and Singapore, that will be a panel data for my analysis and the distance will not be a constant in that case.
Is that right?

-Raisa-

[Joao Santos Silva]

Exactly :-)

[Raisa Yuliana]

Thank you so much, Joao. I think i can run my analysis now

[Raisa Yuliana]

Joao,
I want to ask one more question. The suggestion to " compute descriptive statistics for distance to check that it is constant," do you mean to test the probability value for percentage significant or the std deviation like that? Since I don't know which one I need to see to know if distance is constant or not, so I check all the descriptive statistic test, and I still don't know which one I need to see.
Sorry, if I asked too many questions since this is my first run econometry analysis.
Thank you

-Raisa-

[Joao Santos Silva]

It looks like your distance is not constant. That is very strange; please check your data.

Joao

[Raisa Yuliana]

I still use the basic model with add the population from both countries to see if the population give an impact to the export value. As for the distance, I use the distance from Jakarta main harbor to China. All of the variables I change to natural logarithm first. I attach my raw data if you want to check.
-Raisa-