Dear statalisters,

I have run three similar regressions of this model:

lnGCFGDPit= ß0+ ß1newdealit+ß2lnDIFFintit+ ß3lnDIFFgdpit+ ß4lnFOit+ ß5lnRoGDPit + ß6lnPRODit + ß7lnINFit+ uit (re and robust)

The dependent variable here represents the gross capital flows of nations relative to GDP. I have also run two variants of this model with as the dependent variables the capital inflow of nations relative to GDP and the capital outflow of nations relative to GDP. The absolute values (meaning not relative to GDP) of the in and outflow summed up are the gross capital flow in my model (As I have specified the in and outflows this is correct). However, the results produced showed that the coefficient of newdeal was 0.41, meaning that a newdeal would lead to an 0.41% in the gross capital flows (significant at the 5% level). However, this particular coefficient when looking at capital inflows is 0.36 (significant at the 10% level) and the coefficient for capital outflows as the dependent variable is 0.56 (significant at the 5%level).

In my opinion the coefficients of the in and outflow summed up should be equal to the coefficient of the newdeal variable when it has gross capital flows as its dependent variable due tot the fact that the absolute values of the in and outflows summed up are equal to the absolute value of the gross capital flow. The model run for the three regression is the same except for the difference in dependent variable. Is my intuition correct? Should they add up? And if so, why don't they?

Thank you in advance for your time!