(Beginner!) Questions about regression on CD album sales/internet use/digital music s

#1
Hello,

I am attempting to perform a regression on what has caused the reduction in CD album sales, and need a little assistance. I have 28 data points from the USA 1983-2010, and am using gdp per capita, internet use, digital album sales, and digital single sales as independent variables.

Is it as simple as this? [Eviews]


I can see there that I have a very high R squared, I have heard this could point to some multicollinearity? Is this true?

Also, I predict that internet use not only reduces cd album sales, but increases digital music sales. Could this hint at why the coefficient and t-stat is much higher for hi-speed-inet use, because the regression software has worked this out for me?

Also, in this case what does the negative coefficient represent? I don’t understand how zero income, internet use, etc could cause MINUS cd sales!

Thanks for any help or ideas, really appreciated!
 
#2
Re: (Beginner!) Questions about regression on CD album sales/internet use/digital mus

I guess all the variables in your regression are trending variables (positive trend for all the indep. vars and neg. trend for the dep. var), and this probably explains the high R-squared, as the trend "explains" almost everything.
You could try adding time as an regressor and see if that changes anything. Or perhaps better: regress change in cd sales on change of gdp, change in highspeed internet etc, i.e. cdsales(t - t-1) as dependent variable.

There is no need to worry about the negative constant. There is no way that GDP/capita in the US can be zero, so the coefficient doesnt really make any sense in this specific regression.