Non-linear model using exp()

I have a simple question about the transformation of the variables from levels to logs.
Lets say that the original model has the following goods market clearing condition: Y1+Y2=C1+C2+I1+I2;
If I write:

model; Y1+Y2=C1+C2+I1+I2; end;
The IRF will show absolute deviations from steady-state.

IF I wanted to obtained log-deviations, then I need to transform each variable with exp().
I have seen many models using the following transformation for this very specific equation ( which is originaly linear)

model; exp(Y1)+exp(Y2)=exp(C1)+exp(C2)+exp(I1)+exp(I2); end;
BUT, the exp(Y1+Y2)=exp(Y1)*exp(Y2), then the above transformation is wrong?
How should I do this transformation?

Many thanks in advance.

Sorry, but I don’t understand your question. Every single variable is log-transformed, not the whole equation. Please see Pfeifer(2013): “A Guide to Specifying Observation Equations for the Estimation of DSGE Models”