Dear all,
I am trying to estimate the effect of local competitiveness on internal migration using a gravity model and OLS. I have cross-sectional data for 80 origins and 80 destinations (only one observation for each origin-destination pair). Second, I am using instrumental variable to address the endogeneity problem of my first specification, which only has the typical variables of a gravity model, a contiguity dummy, and two competition variables for origin and destination.
My questions arise regarding the multilateral resistance to migration. Following "Bertoli, S., and J. Fernandez-Huertas Moraga (2015), ‘The Size of the Cliff at the Border,’RegionalScience and Urban Economics,51, March, 1–6", I should include origin or destination dummies, since including both would absorb my two variables of interest, i.e. competitiveness in the origin and competitiveness in the destination. Although it is not clear to me how the variables are absorbed since when running the regression with both origin and destination dummies, the results do not delete any of my variables of interest, although I have multicollinearity with 6 of the dummies (3 from origin and 3 from destination). Although when I use the reghdfe or areg commands I do see how my competitiveness and population variables are omitted from the results. The code I have used is the following:
Or:
The second problem I have is that when I use instrumental variable and include origin dummies, I obtain multicollinearity between those dummies and my instruments, so Stata doesn't give me any results.
In this case should I use ivreghdfe instead of ivregress?
I would very much appreciate any comments on the best way to approach multilateral resistance to migration given my cross-sectional data, and possible reasons/solutions to multicollinearity both using OLS and especially with IV.
Note: I am using OLS and not PPML, because my advisor requested me to use it as part of the estimations.
I am trying to estimate the effect of local competitiveness on internal migration using a gravity model and OLS. I have cross-sectional data for 80 origins and 80 destinations (only one observation for each origin-destination pair). Second, I am using instrumental variable to address the endogeneity problem of my first specification, which only has the typical variables of a gravity model, a contiguity dummy, and two competition variables for origin and destination.
My questions arise regarding the multilateral resistance to migration. Following "Bertoli, S., and J. Fernandez-Huertas Moraga (2015), ‘The Size of the Cliff at the Border,’RegionalScience and Urban Economics,51, March, 1–6", I should include origin or destination dummies, since including both would absorb my two variables of interest, i.e. competitiveness in the origin and competitiveness in the destination. Although it is not clear to me how the variables are absorbed since when running the regression with both origin and destination dummies, the results do not delete any of my variables of interest, although I have multicollinearity with 6 of the dummies (3 from origin and 3 from destination). Although when I use the reghdfe or areg commands I do see how my competitiveness and population variables are omitted from the results. The code I have used is the following:
Code:
reg loutmi licc_origen licc_destination lpoporigen lpopdestination ldist contig i.origin, robust reghdfe loutmi licc_origen licc_destination lpoporigen lpopdestination ldist contig, absorb(origen) vce(robust)
Code:
areg loutmi licc_origen licc_destination lpoporigen lpopdestination ldist contig i.origin, absorb(origen) vce(robust)
Code:
ivregress 2sls loutmi lpoporigen lpopdestination ldist contig (licc_origen licc_destination = ligm_origen ligm_destination) i.origin, first vce(robust) note: 3.origen identifies no observations in the sample note: 24.origen identifies no observations in the sample note: 80.origen omitted because of collinearity note: ligm_origen dropped due to collinearity equation not identified; must have at least as many instruments not in the regression as there are instrumented variables r(481);
In this case should I use ivreghdfe instead of ivregress?
I would very much appreciate any comments on the best way to approach multilateral resistance to migration given my cross-sectional data, and possible reasons/solutions to multicollinearity both using OLS and especially with IV.
Note: I am using OLS and not PPML, because my advisor requested me to use it as part of the estimations.
