Hey 
I am writing my thesis currently about the effects of bilateral aid, trade and fdi on bilateral migration. My dependent variables are migration stock % of origin population and migration stock % of destination population; dependent variables are trade in goods and exports(exports from origin country to destination country), FDI flow (flowing into origin from destination country) and aid (flowing into origin country from destination country). In the last part, I run a two-step GMM estimation to account for serial correlation and endogeneity. I believe that migration might affect trade, FDI and, most likely, aid.
I question whether it's bad to use total trade to all countries (total exports from the origin country) and FDI (total inflows from all countries into the origin country) as lagged instruments, not the dependent variables. Its the only way that the Hansen test has turned out not statistically significant. here is my stata code:
xtabond2 L(0/1).ln_migration_stock_pop_o ln_trade_service_flow_1 ln_fdi_flow_1 ln_aid_flow_1, ///
gmm(ln_migration_stock_pop_o, collapse laglimits(2 3)) ///
gmm(ln_trade_service_flow_1_total_o ln_fdi_flow_1_total_o, collapse laglimits(1 1)) ///
iv(i.year, equation(level)) ///
twostep robust small
I also find that the results are sensitive to whether I use the independent variables in level values or thousands of USD. Any help or recommendations would be greatly appreciated!
I am writing my thesis currently about the effects of bilateral aid, trade and fdi on bilateral migration. My dependent variables are migration stock % of origin population and migration stock % of destination population; dependent variables are trade in goods and exports(exports from origin country to destination country), FDI flow (flowing into origin from destination country) and aid (flowing into origin country from destination country). In the last part, I run a two-step GMM estimation to account for serial correlation and endogeneity. I believe that migration might affect trade, FDI and, most likely, aid.
I question whether it's bad to use total trade to all countries (total exports from the origin country) and FDI (total inflows from all countries into the origin country) as lagged instruments, not the dependent variables. Its the only way that the Hansen test has turned out not statistically significant. here is my stata code:
xtabond2 L(0/1).ln_migration_stock_pop_o ln_trade_service_flow_1 ln_fdi_flow_1 ln_aid_flow_1, ///
gmm(ln_migration_stock_pop_o, collapse laglimits(2 3)) ///
gmm(ln_trade_service_flow_1_total_o ln_fdi_flow_1_total_o, collapse laglimits(1 1)) ///
iv(i.year, equation(level)) ///
twostep robust small
I also find that the results are sensitive to whether I use the independent variables in level values or thousands of USD. Any help or recommendations would be greatly appreciated!
