I have panel data of 25 countries and 20 years. I am implementing ivreg2, gmm2s command to apply IV-GMM in my analysis. I want to know how conflict across countries affects trade. I use some policy and innovation variables (domestic and international policy and innovation) and want to know how they moderate this relationship. I am also using some control variables such as freedom to trade, human development, economic complexity and financial development. I suspect endogenety and instrument all the independent variables by their 2 lags. The command is as follows:
...
ivreg2 lntrade (lnconflict lndomesticpatent internationalpatent lndomesticpolicy lninternationalpolicy lndemocracy lntradefreedom lnfinancialdevelopment economic_complexity lnhuman_development= l(1/2).lnconflict l(1/2).lndomesticpaten l(1/2).internationalpatent l(1/2).lndomesticpolicy l(1/2).lninternationalpolicy l(1/2).lndemocracy l(1/2).lntradefreedom l(1/2).llnfinancialdevelopment l(1/2).economic_complexity l(1/2).lnhuman_development), gmm2s first robust
...
My questions are:
1) Is my command correct?
2) Am i doing it right considering all variables as endogenous? I know ultimately this is the question of economic theory, but i wanted to know in general if this is okay to use two lags (please do not focus on how many lags i am using for now) as internal instruments for all the independent variables in my list? or should I just stick to the ones I am interested in the most ((lnconflict lndomesticpatent internationalpatent lndomesticpolicy lninternationalpolicy)? or should i stick to only conflict variable and consider others as exogenous? Is it conventional to use all independent variables as endogenous (similar to xtabond2 command)?.
3) Note that I am treating control variables as endogenous too (for example trade affects the freedom to trade, i.e. reverse causality). In general, do we need to be concerned about the endogeneity of control variables or do we leave them as they are (that is, treat them as control variables, nothing more than that)
4) I wanted to interact between lnconflict and other policy and innovation variables. So i generated several new interaction variables. For example, by multiplying international patent and lnconflict, i generated lnconflictinterpatent variable, which denotes the interaction between lnconflict and international patent. Now i am just worried if my command below is correct:
...
ivreg2 lntrade (lnconflict lndomesticpatent internationalpatent lndomesticpolicy lninternationalpolicy lndemocracy lntradefreedom lnfinancialdevelopment economic_complexity lnhuman_development nconflictinterpatent= l(1/2).lnconflict l(1/2).lndomesticpaten l(1/2).internationalpatent l(1/2).lndomesticpolicy l(1/2).lninternationalpolicy l(1/2).lndemocracy l(1/2).lntradefreedom l(1/2).llnfinancialdevelopment l(1/2).economic_complexity l(1/2).lnhuman_development l(1/2).lnconflictinterpatent), gmm2s first robust
...
Here, i used 2 lags as instrument for the interaction variable as well. Is this the correct way?
5) Also while creating interaction variables, I use the multiplication between lnconflict and other non-logarithmic variables such as internationalpatent. I did not generate the interaction variable like this: gen coflictinternationalpatent=conflict*internationalp atent and then taking log of it. Rather I used the following command: lnconflictinternationalpatent=lnconflict*internati onalpatent, is this correct?
6) In terms of interaction between two logarithmic variables (lnconflict and lndomesticpolicy), I used:
gen lnconflictdomesticpolicy=lnconflict*lndomesticpoli cy
is this correct? or should I have done this:
gen conflictdomesticpolicy=conflict*domesticpolicy
gen lnconflictdomesticpolicy=ln(conflictdomesticpolicy )?
I apologize for so many questions at once,
...
ivreg2 lntrade (lnconflict lndomesticpatent internationalpatent lndomesticpolicy lninternationalpolicy lndemocracy lntradefreedom lnfinancialdevelopment economic_complexity lnhuman_development= l(1/2).lnconflict l(1/2).lndomesticpaten l(1/2).internationalpatent l(1/2).lndomesticpolicy l(1/2).lninternationalpolicy l(1/2).lndemocracy l(1/2).lntradefreedom l(1/2).llnfinancialdevelopment l(1/2).economic_complexity l(1/2).lnhuman_development), gmm2s first robust
...
My questions are:
1) Is my command correct?
2) Am i doing it right considering all variables as endogenous? I know ultimately this is the question of economic theory, but i wanted to know in general if this is okay to use two lags (please do not focus on how many lags i am using for now) as internal instruments for all the independent variables in my list? or should I just stick to the ones I am interested in the most ((lnconflict lndomesticpatent internationalpatent lndomesticpolicy lninternationalpolicy)? or should i stick to only conflict variable and consider others as exogenous? Is it conventional to use all independent variables as endogenous (similar to xtabond2 command)?.
3) Note that I am treating control variables as endogenous too (for example trade affects the freedom to trade, i.e. reverse causality). In general, do we need to be concerned about the endogeneity of control variables or do we leave them as they are (that is, treat them as control variables, nothing more than that)
4) I wanted to interact between lnconflict and other policy and innovation variables. So i generated several new interaction variables. For example, by multiplying international patent and lnconflict, i generated lnconflictinterpatent variable, which denotes the interaction between lnconflict and international patent. Now i am just worried if my command below is correct:
...
ivreg2 lntrade (lnconflict lndomesticpatent internationalpatent lndomesticpolicy lninternationalpolicy lndemocracy lntradefreedom lnfinancialdevelopment economic_complexity lnhuman_development nconflictinterpatent= l(1/2).lnconflict l(1/2).lndomesticpaten l(1/2).internationalpatent l(1/2).lndomesticpolicy l(1/2).lninternationalpolicy l(1/2).lndemocracy l(1/2).lntradefreedom l(1/2).llnfinancialdevelopment l(1/2).economic_complexity l(1/2).lnhuman_development l(1/2).lnconflictinterpatent), gmm2s first robust
...
Here, i used 2 lags as instrument for the interaction variable as well. Is this the correct way?
5) Also while creating interaction variables, I use the multiplication between lnconflict and other non-logarithmic variables such as internationalpatent. I did not generate the interaction variable like this: gen coflictinternationalpatent=conflict*internationalp atent and then taking log of it. Rather I used the following command: lnconflictinternationalpatent=lnconflict*internati onalpatent, is this correct?
6) In terms of interaction between two logarithmic variables (lnconflict and lndomesticpolicy), I used:
gen lnconflictdomesticpolicy=lnconflict*lndomesticpoli cy
is this correct? or should I have done this:
gen conflictdomesticpolicy=conflict*domesticpolicy
gen lnconflictdomesticpolicy=ln(conflictdomesticpolicy )?
I apologize for so many questions at once,
