Hi Statalist,

Whilst this is not a purely stata related question, it relates more to econometrics, and am posting here because I cannot find the answer anywhere on the internet.

In a regression where the interaction term is between a log-transformed variable and a dummy variable, how should that coefficient be interpreted? Here, D1 is a dummy variable, and y is a level variable

e.g. y=α+β1D1+β2ln(X)+δ1(ln(X)∗D1)

I'm looking to find the additional effect on y from a 1% rise in X for D1=1 vs D1=0.

Would the answer be:

a) (exp(δ1)-1)/100

or

b) (exp(β2+δ1)-1)/100 minus (exp(β2)-1)/100