OLS Regression, dummy variables issue

[edj1991]

Hi,

I am trying to estimate a simple growth regression to try to find a positive impact of mobile penetration on average growth rates in african countries. I can estimate the OLS however I am trying to include dummy variables for each country (-1 for reference). When I include all variables plus dummies for each country EVIEWS gives me the message 'insufficient number of observations'. Can anyone shed some light/offer some advice as to why this is happening? As when I run the regression without dummies I can get an output.

Thanks

[startz]

You can't have a dummy for each country when you only have one observation for each country.

[edj1991]

so dummy variables being used here is useless? I was doing this based on advice from a tutor. Perhaps she did not understand my specification.

[startz]

Not useless, impossible.

[edj1991]

However dummies for say country regions (eg. northern africa) would work here?

[startz]

Mechanically, yes that would work. You might ask why dummy variables should be used here. For that matter, you might ask why cell phone penetration causes GDP growth rather than the other way around.

[edj1991]

as there may be significant differences amongst different nations within Africa, as some as upper middle income countries whereas many are low income countries. Yes there is the possibility/likelihood of endogeneity, reverse causality. My problem is that I have tried to replicated a similar specification to those used in the relevant literature however my coefficient for mobile penetration is coming out as negative.