Dear all of Statalist,
I am applying a gravity model with origin-time FE, destination-time FE and origin-destination FE. My interest variable is origin-destination-time. In this case, I am already applying Baier and Bergstrand (2007) that recommends that the inclusion of country-pair FE is an appropriate econometric method of accounting for reverse causality issues. To have a strictest exogeneity test, I have applied what suggested Wooldridge (2010, chapter 10). I have added the future lead of my interest variable and, under strict exogeneity, the parameter associated with this variable should be statistically insignificant. It is.
This is what I have done.
But, I have seen that there are a lot of papers that use instead of lead the lag variable. I cannot see the difference between them. More than understand the econometric issue, I would like to understand the sense background.
Those are the possibilities:
Trade = FE_it + FE_jt + FE_ij + Conectivity + Conectivity_t+1 = it is exogenous if the Conectivity_t+1 is not correlated with the current trade. This is what I did and my interest variable is exogenous.
Trade = FE_it + FE_jt + FE_ij + Conectivity + Conectivity_t-1 = the current trade is not reflection of the past connectity. This is what I see in a lot papers.
What I have to consider? They check different thinks?
Thank you in advance,
Ticiana
I am applying a gravity model with origin-time FE, destination-time FE and origin-destination FE. My interest variable is origin-destination-time. In this case, I am already applying Baier and Bergstrand (2007) that recommends that the inclusion of country-pair FE is an appropriate econometric method of accounting for reverse causality issues. To have a strictest exogeneity test, I have applied what suggested Wooldridge (2010, chapter 10). I have added the future lead of my interest variable and, under strict exogeneity, the parameter associated with this variable should be statistically insignificant. It is.
This is what I have done.
But, I have seen that there are a lot of papers that use instead of lead the lag variable. I cannot see the difference between them. More than understand the econometric issue, I would like to understand the sense background.
Those are the possibilities:
Trade = FE_it + FE_jt + FE_ij + Conectivity + Conectivity_t+1 = it is exogenous if the Conectivity_t+1 is not correlated with the current trade. This is what I did and my interest variable is exogenous.
Trade = FE_it + FE_jt + FE_ij + Conectivity + Conectivity_t-1 = the current trade is not reflection of the past connectity. This is what I see in a lot papers.
What I have to consider? They check different thinks?
Thank you in advance,
Ticiana
