Model in log

[bigbigben]

I have two versions of the same model, one is with variable in log and the other one is with variable in level. The one in level works fine but the one with log is not. I just wonder whether it is effecient to rewrite some equations in log to the linear form, say, orginally, I have

exp(Y)=exp(K)^alpha*exp(L)^(1-alpha);

Is the following more efficient, given there is another equation exp(Y)=exp(C)+exp(I);

Y=alpha*K+(1-alpha)*L;

[MichelJuillard]

One version isn't more efficient than another. Sometimes a linear approximation will be more accurate, sometimes a loglinear one.

For growth model, it is better to use a loglinear approximation.

Usually, one can't use the same initval values with a linear or a loglinear approximation as in one case the variables are in level, in the other one is computing the log of the variables at the steady state.

Kind regards

Michel