# Interpreting regression coefficents when the variables are in proportions

I have the following cross-sectional regression.

per capita car ownership = 0.025*per capita college degrees + 0.012*availability of underground(dummy variable) - 0.287

I am having difficulty with interpreting the coefficients of both per capita car ownership and per capita college degrees. Since they are percentages, should I treat them like log-log level and interpret like 1% increase in per capita college degrees is likely to increase per capita car ownership by 0.025%?

First, car ownership should not have a parameter associated with it; as you have written the model, it is a dependent variable (and, indeed, you do not have a coefficient for it).

The model means that, if per capita degrees goes up by 1 unit, it is predicted that per capita car ownership will go up by .025 (ignoring the dummy variable for simplicity). How each of these is measured isn't clear; literally, "per capita" would mean "per person" (actually, per head), but it may be that these are measured per thousand or something.