Hi All,

I'm running a panel regression, with my dependent variable being the foreign value-added share of domestic exports. I have 64 countries and 11 time periods (2005-15). I'm running two separate regressions: the first is with Y at the aggregate level, and the second is with Y at the manufacturing level. The independent variables remain the same in both cases, since they are presumed to affect both the total and manufacturing Y similarly. My RHS is a mixture of share variables, levels, indices (such as overall index of regulatory quality) and a 2009 dummy. I'm using Stata/SE 14.2.

I have two questions. I'm using -xtreg fe (country fixed-effects) and clustering at the country level:

1. How do I ensure that my RHS falls between 0 and 1, since my dependent is a share variable?
2. When I change my Y from the aggregate indicator to the manufacturing level, the results change radically. Almost all the variables seem to lose significance, even though both aggregate Y and manufacturing Y are highly positively correlated and display the same pattern over 2005-15. Are there any specific circumstances under which this might happen? Are there any diagnostic tests that might help me understand this problem better?


Regards,
Saunok Chakrabarty