Dear Sir/ Madam,

I am doing a research using gravity model of trade about the influences of ACFTA (ASEAN China Free Trade Area) on Vietnam’s export to ACFTA’ members. My probem with the estimation results is that most of my coefficiencts’ significance of explanatory variable are at 10% level of confidence, this is a major setback for me as I believe that a number of explanatory variables (such as GDP, GDP per Capita, ...) should have higher level significance.

Please allow me to represent the database, the regression model, the estimation methods and the stata command that I have used so far. I would really be grateful if you can point out where I have done wrong and where I can correct my mistakes.

1. First is the database (attached in this forum below in excel and dta. File). The database consists of 16 countries (Vietnam as the main exporter and the rest 15 countries are Vietnam’s export destinations), 15 pairs across 19 years (2000 – 2018), 285 observations in total. The data will be transform to longidunital panel data in Stata.

2. Secondly, its the gravity model regression and estimation methods

The gravity model regression I choose to analyse the relations between trade (dependent variables) and other explanatory variables as follow:

xtreg ln_trade ln_gdp_exp ln_gdp_imp ln_gdppc_exp ln_gdppc_imp ln_distw contig fta_vjfta fta_vkfta fta_vanzfta fta_vifta fta_acfta, re

The ln stands for natural logarithm from the trade, gdp, gdppc (gdp per capita), distance. The rest are dummy variables with the value of 0 or 1.

I did run both Fixed effect model and Random effect model, the results of Hausman test suggest Random effect model being better. Please see my estimation results of random effect model in the attached jpeg file 📷

As you can see, the problem is that major explanatory variables has significance at 10% level, which is highly prone to mistake and rejected with 5% level of confidences. Furthermore, the ln_distance has positive relations with dependent variable (it should have been negative sign in my knowledge).

I would be really grateful if you can tell me where I have made mistakes, it is the database or the estimation methods that are wrong, or something else.

Thank you for your precious assistance, I wish the best for all.